Can computer models help us to understand human creativity?

Creativity and computers: what could these possibly have to do with one another? “Nothing!,” many people would say. The two are simply incompatible.”

Well, I disagree. Computers and creativity make interesting partners with respect to two different projects. One, which interests me the most, is understanding human creativity. The other is trying to produce machine creativity–or anyway, machine “creativity”–in which the computer at least appears to be creative, to some degree.

What is Creativity?

Human creativity is something of a mystery, not to say a paradox. One new idea may be creative, while another is merely new. What’s the difference? And how is creativity possible? Creative ideas are unpredictable. Sometimes, they even seem to be impossible — and yet they happen. How can that be explained?

Before we can hope to explain creativity, we need to know what’s meant by the term. In fact, people use it in rather different ways — so, when discussing it, they can end up talking at cross purposes.

Here, let’s agree that creativity is the ability to come up with ideas or artefacts that are new, surprising, and valuable. “Ideas,” here, includes concepts, poems, musical compositions, scientific theories, cooking recipes, choreography, jokes … and so on. “Artefacts” include paintings, sculpture, steam-engines, vacuum cleaners, pottery, origami, penny-whistles … and more.

Creativity isn’t a special “faculty,” confined to a tiny elite: it’s an aspect of human intelligence in general. Nor is it an all-or-none affair. Rather than asking “Is that idea creative, Yes or No?,” we should ask “Just how creative is it, and in just which way(s)?” Asking that question will help us to see just what sorts of psychological process could have brought the new idea about.

Creative ideas, then, are new. But of course, there’s new–and there’s new. Ask a teacher, for instance. Children can come up with ideas that are new to them, even though they may have been in the textbooks for years. Someone who comes up with a bright idea is not necessarily less creative just because someone else had it before them.

We need to make a distinction between “psychological” creativity and “historical” creativity. (P-creativity and H-creativity, for short.) P-creativity is coming up with a surprising, valuable idea that’s new to the person who comes up with it. It doesn’t matter how many people have had that idea before. But if a new idea is H-creative, that means that (so far as we know) no-one else has had it before: it has arisen for the first time in human history.

Clearly, H-creativity is a special case of P-creativity. For historians of art, science, and technology, H-creativity is what’s important. But for someone who is trying to understand the psychology of creativity, it’s P-creativity which is crucial. Never mind who thought of the idea first: how could anyone manage to come up with it, given that they had never thought of it before?

If “new,” in this context, has two importantly different meanings, “surprising” has three. First, An idea may be surprising because it’s unfamiliar, or even unlikely–like an outsider winning the Derby. This sort of surprise goes against statistics.

The second sort of surprise is more interesting. An unexpected idea may “fit” into a style of thinking that you already had–but you’re surprised because you hadn’t realized that this particular idea was part of it.

And the third sort of surprise is more interesting still. This is the astonishment you feel on encountering an apparently impossible idea. It just couldn’t have entered anyone’s head, you feel–and yet it did. What on earth can be going on?

The Three Roads to Creative Surprise

“What is going on” isn’t magic–and it’s different in each type of case. For creativity can happen in three main ways, which correspond to the three sorts of surprise.

The first involves making unfamiliar combinations of familiar ideas. Examples include poetic imagery, collage in painting or textile art, and analogies. Think of a physicist comparing an atom to the solar system, for instance, or call to mind some examples of creative associations in poetry or in political cartoons.

In all these cases, making–and appreciating–the novel combination requires a rich store of knowledge in the person’s mind, and many different ways of moving around within it. If the new combination is to be valued, it has to have some point. It may or (more usually) may not have been caused by some random process–like shaking marbles in a bag. But the ideas/marbles have to have some intelligible conceptual pathway between them for the combination to “make sense.”

The other two types of creativity are interestingly different from the first. They involve the exploration, and in the most surprising cases the transformation, of conceptual spaces in people’s minds.

Conceptual spaces are structured styles of thought. They aren’t originated by one individual mind, but are picked up from one’s culture, or occasionally borrowed from other cultures. They include ways of writing prose or poetry; styles of sculpture, painting, or music; theories in chemistry or biology; fashions of couture or cooking … in short, any disciplined way of thinking that’s familiar to (and valued by) a certain social group.

Within a given conceptual space, many thoughts are possible, only some of which may have been actually thought. Some spaces have a richer potential than others. Noughts-and-crosses is such a restricted style of game-playing that every possible move has already been made countless times. But that’s not true of chess, where the number of possible moves, though finite, is astronomically large. So is the space of possible sonnets, or screenplays, or fugues.

Someone who comes up with a new idea within a particular thinking-style is being creative in the second, exploratory, sense. If the new idea leads on to others (still within the same space) whose possibility was previously unsuspected, so much the better. Exploratory creativity is valuable partly because it can enable someone to see possibilities they hadn’t glimpsed before. They may even start to ask just what limits, and just what potential, this style of thinking has.

All professional artists and scientists do this sort of thing. Even the most mundane street artists produce new portraits every day. They are exploring their space, though not necessarily in an adventurous way. Occasionally, they may realize that their sketching-style enables them to do something (convey the set of the head, or the hint of a smile) better than they’d been doing before. They add a new trick to their repertoire, but in a real sense it’s something that “fits” their established style: the potential was always there.

What the street-artist–or Picasso, in a similar position–may also do is realize the limitations of their style. Then, they have an opportunity to change it.

The limits of the thinking-style, or of some particular aspect of it, may be slightly pushed, slightly altered, gently tweaked. They may even be changed so decisively that ideas which previously were unthinkable now become possible. The deepest cases of creativity involve someone’s thinking something which, with respect to the conceptual spaces in their minds, they couldn’t have thought before. The supposedly impossible idea can come about only if the creator transforms the pre-existing style in some radical way.

But how can that possibly happen? And how could computers help us to find the answer?

How Can Computers Throw Light on Creativity?

To understand how exploratory or transformational creativity can happen, we must know what conceptual spaces are, and what sorts of mental processes could explore and modify them.

Styles of thinking are studied by literary critics, musicologists, and historians of art, fashion, and science. And they are appreciated by us all. But intuitive appreciation, and even lifelong scholarship, may not make their structure clear. (An architectural historian, for instance, said of Frank Lloyd Wright’s Prairie Houses that their “principle of unity” is “occult”.)

This is the first point where computers are relevant. Conceptual spaces, and ways of exploring and transforming them, can be described by concepts drawn from artificial intelligence (AI).

AI tries to get computers to do the many different sorts of things that minds can do. Indeed, AI-concepts enable us to do psychology in a new way, by allowing us to construct (and test) hypotheses about the structures and processes that may be involved in thought. For instance, the structure of tonal harmony, or the “grammar” of Prairie Houses (no longer “occult”), can be clearly expressed, and specific ways of exploring the space can be tried out. Methods for navigating, and changing, highly-structured spaces can be compared.

Of course, there is always the additional question of whether the suggested structures and processes are actually implemented in people’s heads. And that question isn’t always easy to answer. But the point, here, is that a computational approach gives us a way of formulating clear scientific hypotheses about the rich subtleties of the human mind.

With respect to understanding creativity, computer models can help us because they can be creative. Or rather, they can at least appear to be creative.

Many people would argue that no computer could possibly be genuinely creative, no matter what its performance was like. It might produce theories as ground-breaking as Einstein’s, or music as highly valued as Beethoven’s … but still, for these people, it wouldn’t really be creative.

Several different arguments are commonly given. For instance: it’s the programmer’s creativity that’s at work here, not the machine’s. The machine isn’t conscious, and has no desires or values–so it can’t appreciate or judge what it’s doing. A work of art is an expression/communication of human experience, so machines simply don’t count. And all ideas have meaning, which is lacking in computers.

Perhaps you accept at least one of those reasons for denying creativity to computers? Very well, I won’t argue with you here. Let’s assume, for the purpose of this discussion, that computers can’t really be creative. That doesn’t mean, however, that there’s nothing more of interest to say.

All the objections just listed accept, for the sake of argument, that the imaginary computer’s performance is indeed very like that of human beings. What we need to focus on here is whether it’s true that computers could, in fact, come up with ideas that at least appear to be creative–and, if so, how?

Computer Models of Creativity

Let’s consider combinational creativity first. In one sense, this is easy to model on a computer. For nothing is simpler than picking out two ideas (two data-structures) and putting them alongside each other. A computer could merrily produce novel combinations till Kingdom come.

But would they be of any interest? We saw, above, that combining ideas creatively isn’t like shaking marbles in a bag. There must be some intelligible, though previously unnoticed, link between them that we value because it is interesting in some way. We saw also that combinational creativity typically requires a very rich store of knowledge, and the ability to form links of many different types.

For a computer to make a subtle combinational joke, for example, would require (1) a database with a richness comparable to ours, and (2) methods of link-making (and link-evaluating) comparable in subtlety with ours. In principle, this isn’t impossible. After all, the human mind/brain doesn’t do it by magic. But don’t hold your breath!

The best example of computer-based combinational creativity so far is a program called JAPE, which makes punning jokes of nine general types that are familiar to every ten-year-old. For example: What do you call a depressed train?–A low-comotive, and What’s the difference between leaves and a car?–One you brush and rake, the other you rush and brake. To be able to do this, the program needs a set of templates defining the ’skeleton’ of each type of joke (e.g. What’s the difference between an x and a y?, What kind of x can y?, and What do you get when you cross an x with a y?), plus rules for finding words to fit the chosen template. Those rules, in turn, need access to a large semantic network (of over 30,000 items), with links representing not only meaning, hierarchy, and synonymy but also phonology, spelling, syllabic structure, and grammatical class.

Filling-in a familiar joke-schema is difficult enough. (Try to work out just what was needed to generate the joke about the depressed train.) But making a one-off jest is usually more demanding. Ask yourself, for instance, what Jane Austen had to know in order to write the opening sentence of Pride and Prejudice: “It is a truth universally acknowledged that a single man in possession of a good fortune must be in want of a wife.” (And why, exactly, is it funny?) To put the relevant knowledge into a computer, alongside (so as not to ’cheat’) the many other things that Austen happened to know, would take forever. And to enable the program to originate the countless one-off jokes in the book (in Mr. Collins’ preposterous proposal to Elizabeth, for instance) would in practice be impossible.

In short, computer models of combinational creativity can help us to understand, in general terms, how our own combinations can come about–but they will generate valuable new combinations only rarely.

Exploratory creativity is more promising. Indeed, several programs already exist which can explore a given space in acceptable ways.

One example is Harold Cohen’s AARON, a drawing-program that can generate thousands of line-drawings or coloured images in a certain style. These are pleasing enough to be exhibited in galleries around the world. Another is David Cope’s “Emmy”, which composes music in many different styles (based on human composers such as Bach, Mozart, Stravinsky, and Joplin). Still others include architectural programs that design Palladian villas or Prairie Houses, and programs that can analyse experimental data and find new ways of expressing scientific laws.

A few AI-programs can even transform their conceptual space, by altering their own rules. “Evolutionary” programs, for instance, can make random changes in their current rules so that new forms of structure result. At each generation, the “best” structures are selected, and used to breed the next generation. Several examples evolve coloured images which, like AARON’s, are exhibited in galleries world-wide. These images often cause the third, deepest, form of surprise. In such cases, one can’t see the relation between the daughter-image and its parent. The one appears to be a radical transformation of the other, or even something entirely different.

Values and Creativity

There’s no major difficulty in getting an (evolutionary) art program to make transformations: that’s relatively easy. What’s difficult is to state our aesthetic values clearly enough to enable the program itself to make the evaluation at each generation. At present, the “natural selection” is done by a human being. (In scientific domains, the value-criteria can often be stated clearly enough to allow the program to apply them automatically. So these techniques are used, for instance, to help biochemists to design new molecules for pharmaceutics.)

One huge problem here has no special relevance to computers, but bedevils discussion of human creativity too. I said earlier that “new” has two meanings, and that “surprising” has three. I didn’t say how many meanings “valuable” has–and nobody could. Our aesthetic values are difficult to recognize, more difficult to put into words, and even more difficult to state really clearly. (For a computer model, of course, they have to be stated really, really clearly.)

Moreover, they change. They vary across cultures. They are often disputed: different subcultures or peer groups value different types of dress, jewellery, or music. And where transformational creativity is concerned, the shock of the new may be so great that even fellow artists or fellow-scientists find it difficult to see value in the novel idea.

Because creativity by definition involves not only novelty but value, and because values are highly variable, it follows that many arguments about creativity are rooted in disagreements about value. This applies to human activities no less than to computer performance. So even if we could identify and program our aesthetic values, so as to enable the computer to inform and monitor its own activities accordingly, there would still be disagreement about whether the computer even appeared to be creative.

The answer to our opening question, then, is that there are many intriguing relations between creativity and computers. Computers can come up with new ideas, and help people to do so. And computer models of creativity, both in their failures and in their successes, help us think more clearly about our own creative powers.

23 comments to Can computer models help us to understand human creativity?

  • Mark

    I would like to add a further distinction to creativity that addresses the method through which a creative idea or artifact was produced.

    Margaret Boden’s three routes to “creative surprise” define three types of creativity: combinational, exploratory, and transformational. But there is further division regarding how an agent, either artificial or natural, can go about producing creative ideas or artifacts in those three veins. The first technique is exhaustion or brute-force of all ideas or artifacts within a conceptual space, noting that conceptual transformations are their own conceptual space as well, and the second technique is instinctual pruning of creative attempts. Following Boden’s cue, I will call these BF-creativity and I-creativity for brute-force creativity and instinctual creativity, respectively.

    Commonly cited BF-creativity examples include Edison’s creation of the light bulb by trying numerous materials before chancing across a suitable filament. Such creativity did not so much result from insight as it did from determination as Edison and his team set about exhausting all possible light bulb filaments. The key to such BF-creativity is the value metric which distinguishes “creative” results from “non-creative” results. When presented with a metric such as “Find a material suitable as a light bulb filament,” a BF-creative entity will exhaust all possible artifacts in the conceptual space, namely materials, until it arrives at one the desired set of properties.

    I-creativity has much more subtlety, however. The model for the I-creative entity is the experienced research scientist or the skilled author who instinctively or intuitively knows how to work creatively. I-creative entities prune the conceptual space and explore only a limited set of options, thereby more effectively focusing their efforts. Historical examples suggest that I-creativity is innate, as in the case of child prodigies, or is learned with much experience and practice, like a skilled sculptor.

    Modern artificial intelligences are often BF-creative entities, exhaustively searching for creative artifacts or ideas using an often narrow value metric. I tend to think of chess engines that consider all possible moves, selecting only those that yield the best victory, and I think intelligences like JAPE or AARON are still mostly BF-creative. JAPE is provided with a metric for identifying good jokes and then purposelessly searches for a sequence of satisfactory letters. AARON’s creative effort appears more abstract and is of more technical complexity than JAPE’s but still reduces to patterned or rule-based BF-creativity. Not knowing the inner details of AARON, I might even venture that AARON’s “creativity engine” is deterministic unless supplied with a source of randomness, which seems to prohibit the development of I-creativity.

  • In her article, “Creativity in an Nutshell”, Margaret Boden classifies creativity
    into three types: (1) the unfamiliar combination of familiar ideas; (2)
    the exploration of conceptual spaces; and (3) the transformation of conceptual
    spaces. I will argue that these conceptual spaces are organised into a hierarchy:
    above each conceptual space is a meta-space that describes how the lower one
    may be transformed. From this viewpoint, transformation of the lower space
    can also be seen as exploration of the upper one. This blurs the otherwise sharp
    distinction that Boden draws between creativity of types (2) and (3). This
    viewpoint was partly anticipated in my [Bundy, 1994], an earlier commentary
    on Boden’s theories of creativity.

    To illustrate this two-level viewpoint, consider the game of rugby. I choose
    this particular game because its rules are subject to frequent amendment, but
    any other game could be used to illustrate my point. We can regard the rules
    of rugby as a conceptual space. Exploration in this space aims to discover novel
    ways to play the game within the rules. Such novelty is intended to open up
    new tactical opportunities for your team and to take the opponents by surprise,
    and is creativity of type (2).

    The rules of rugby are, however, being constantly revised. The International
    Rugby Board (IRB) is concerned to find the right balance between making the
    game attractive to spectators and players, and making it safe for players. For
    instance, if it decides that the game is too slow it might introduce rules that
    require rapid recycling of the ball after breakdowns, e.g., rucks. On the other
    hand, if he thinks too many injuries are arising during scrums, then it might
    change the rules of engagement to punish dangerous scrum behaviour, such as
    insufficient binding between the players. Such rule changes can be regarded as
    transforming the conceptual space, i.e., it is creativity of type (3).
    The IRB can also, however, be regarded as engaged in creativity of type
    (2), but within a meta-conceptual space composed of meta-rules, i.e., rules that
    change rules. This meta-space is concerned with issues of attractiveness, safety,
    etc. in rugby.

    The type (3) creativity sometimes consist not just of exploring a well understood
    meta-space, but constructing that meta-space in the first place. the
    culturally isolated musician who first steps outside the narrow confines of the
    musical tradition in which s/he was educated, recognises the wider concerns
    of harmony, melody, consonance, etc., and transforms that tradition, begins to
    build that musical meta-space in which musical traditions are compared and
    generated. Similarly, the architect who understands how traditional buildings
    meet the needs of shelter, security, warmth, etc., but sees alternative ways to
    meet these requirements, is beginning to develop an architectural meta-space.
    This analysis helps explain one of the objections to computational creativity,
    especially of type (3). A computer is usually programmed to transform a
    conceptual space by implementing exploration in the meta-space. Invariably, a
    human programmer has constructed this meta-space, so has done the tough part
    of type (3) creativity, leaving the computer to conduct only type (2) creativity
    by exploring this meta-space.

    Perhaps one hope of overcoming this computational limitation is to create a
    conceptual space that is its own meta-space, i.e., where exploration in the space
    can also transform that same space. This might generate conceptual spaces that
    no human has yet envisaged.

    [Bundy, 1994] Bundy, Alan. (1994). What is the difference between real creativity
    and mere novelty? Brain and Behavioural Sciences,
    17(3):533–34. Also available from Edinburgh as DAI Research
    Paper No. 672.

  • Margaret A. Boden

    Thanks to Mark and Alan for their comments!

    Mark: Although BF-search is always possible in principle, it isn’t possible in practice except for conceptual spaces of a very limited size (e.g. naughts and crosses, or tic-tac-toe). Even the most powerful chess-computers, using specially designed hardware, can only consider a limited number of levels of moves (perhaps 8, or 12, or …)–although admittedly they can consider every move on a given level.

    I don’t now enough about the details of Edison’s invention, but I doubt whether he had a clear enough idea of what might count as “possible” filaments to be able to search them all **exhaustively*. You’re quite right, however, to say that he systematically considered a relatively long list of possibilities before finding “the right one”.

    You’re right when you say that experienced scientists and authors can “intuitively” find creative ideas. But that doesn’t tell us anything, because (as I say in my book) “intuition” is the name of a question, not of an answer. In other words, it is not a term that marks a particular power or skill possessed by the creative person. It’s just a way of saying (1) We manage to do something successfully, and also (2) we don’t know **how** wed o it, and in particular (3) we have no conscious access to how we do it.

    There’s nothing especially awe-inspiring or mystical about this. We don’t have conscious access to how we manage to speak grammatically, either, nor to hoe we manage to identify the objects that we see in front of us. And that’s an evolutionary/computational necessity. It we did have access to all those details, the information overload would be paralyzing.

    JAPE has a set of structured templates, and relevant criteria, which it explores (NOT by BF-search). It doesn’t merely “search for a sequence of satisfactory letters”—unless by that you simply mean that it doesn’t actually **understand** the words that it is using.

    AARON doesn’t use BF-search. It’s not attempting an exhaustive listing of all the possibilities.

    Randomness may or may not contribute to creativity (examples can be given on either side: see Chapter 9 of my book), but the structures, or the **constraints**, are what make creative ideas (a) possible and (b) valuable.

    ALAN: I entirely agree that the E/T distinction is not “sharp”, and that the way I describe it may be misleading for that reason. Your example of the musician pushing forward the space of harmony is a good example. One could say that the whole of post-Renaissance tonal harmony is one space, with no new space arising until atonality, with Schoenberg. Or one could say that moving from single-key melodies to melodies-involving-modulation is exploration, or transformation. But the careful comparison of the various successive spaces/pseudo-spaces helps one to see what is going on, and what is novel/valuable about the new idea–and may even offer hints about **just how** the changes are effected,

    Nice example of the rugby rules. Yes, the meta-space uses rules (attractiveness, safety) set up by the IRB. And they are (more or less) specific to rugby. More general meta-rules may exist too, applicable in virtually any domain. For instance: consider the negative; iterate a procedure; drop a constraint; add/amalgamate constraints…… These were the sort of **general** heuristics that Doug Lenat used in AM.

    Finally, yes: the human programmer normally defines the meta-space as well as the lower-level space. But, arguably, that’s not so if GAs (genetic algorithm) are used, so that the space **evolves**. (I’ve got a chapter in my forthcoming book CREATIVITY AND ART, due out from OUP in September, which discusses that issue in relation to the notion of the artist’s “personal signature”.)

  • Anthony O'Hear

    ‘Can Computer Models Help Us to Understand Human Creativity?’ is the title of Margaret Boden’s interesting and balanced piece. If nothing else Boden draws an important distinction between P-creativity and H-creativity. Failure to mark this distinction bedevils much discussion of education post-Dewey, and much educational practice too, where it seems that P-creativity is at a premium, regardless of the value of what is created. So long as the child is doing something he or she has not done before, we should all applaud. Unfortunately his focus on what is new for the child all too often leads to a neglect in teaching of the essential ground rules for a given form of knowledge or experience, the learning of which may not be particularly creative in either sense. But the absence of this knowledge may then inhibit the possibilities of H-creativity later on, where the assumption is that the creative person will indeed have mastered a body of established technique and knowledge before being able to do something valuable.

    But then, as Boden says, ‘creativity’ has about it a whiff of the valuable. Or at least, the sort of creativity which fills the histories of art and science does. Is all creativity valuable? Perhaps not, if what is meant is (merely) something new. Many new things may be quite trivial, in a broad sense worthless.

    To take us from the merely new to the valuably creative we will have to go beyond the mind of the creator and into the human world of culture, meaning, significance etc., etc. And here I worry that a stress on the workings of the creative mind – whether human or computational – will miss the essential context.

    For the genuinely creative mind has something of value to say, outside of the mind of the creator, valuable to others, to the culture. Can examining the workings of computers, brains, or even of Cartesian minds (on their own) help us to understand the forms of life in which creativity has a role and is valued?

    Further, while I accept that computers can give us models of combinatorial creativity, as Boden says, how far is she confident that these models are replicated in the ways we humans, embedded in a relevant form of life, might make our own mental leaps and bounds, even our puns?

  • Perhaps you accept at least one of those reasons for denying creativity to computers? Very well, I won’t argue with you here. Let’s assume, for the purpose of this discussion, that computers can’t really be creative. That doesn’t mean, however, that there’s nothing more of interest to say.

    I’d like to offer a minor elaboration on this point. Quite independently of whether or not any computer program is or could, in principle, really be creative, until the advent of the digital computer we had no way of developing an explicit and detailed model of any mental process at all. That alone, it seems to me, justifies computational investigation of creativity; it forces to think more deeply and in more detail. By the standards of computer modeling, most other approaches to mental processes are mostly wishful hand-waving.

  • Mary Alice Duffy

    The title of this article “Can Computer Models Help Us To Understand Human Creativity?” confuses me because I think that computers and computer models are a product of human creativity. Through the use of the human mind to conceive ingenious and novel ideas, the concept of a computer came to be. I do not believe that a computer is capable of displaying the same ingenuity or cleverness as a human mind. Whatever capability a computer is able to have, a human is always behind the scenes or responsible for given the computer it’s capabilities. This is because a computer is unable to perform reflective thought. I suppose I would answer this question by saying yes, computer modules can help us to understand human creativity. It is an excellent form of human creativity at its finest especially given the factors mentioned in the article: new, surprising, and valuable. Computers were “historically” new, meaning the idea had not been thought of before. They are surprising because they continue to be unfamiliar and astonishing to many people. Lastly, computers are valuable because they have become considerably beneficial in the lives of humans.

    The main point or statement in the article that supports and agrees with my reasoning is: “With respect to understanding creativity, computer models can help us because they can be creative. Or rather, they can at least appear to be creative.” As I stated earlier I think that computers only simulate creativity because I feel it is a characteristic that requires a mind and behind every computer action is a human hard-wiring or writing the software. The final production from a computer may possess all three qualities of creativity and therefore some people may give credit to the computer itself, however I remain loyal to the thought that the end results of a computer process ultimately reflects on the creativity of the human mind behind it.

  • Margaret A. Boden


    Yes, novelty isn’t enough. And yes, children need practice; they need to learn the styles (rules, conventions, concepts ….) of thinking that are valued in their society if they are to come up with interesting P-creative, or H-creative, ideas.

    Our “forms of life” are reflected in and arise from our memories and conceptual associations, which involve the social group but are implemented in individuals’ heads. In principle, these could be modelled in a computer. But in practice, and in most domains (especially literary domains) this is too richly complex to be so modelled.


    I couldn’t agree with you more!!! Computational concepts and models provide (for the first time) ways of clearly expressing, and even of testing the coherence of, psychological theories about how we think.


    Yes, the human programmer is ultimately responsible, and it is human creativity which has provided us with computers. However, it doesn’t follow that the programmers can always (or even often) predict just what their program will do. That’s especially true when the program can change itself–for instance, by learning (as a result of interactions with the external world), or by evolution (i.e. using genetic algorithms).

  • Maggie Boden’s essay on creativity is a wonderful demonstration of how thinking about what intelligent machines can do, forces us to reflect deeply on the nature of the creative process. This brief meditation is filled with those helpful distinctions that facilitate a better appreciation for the numerous facets of being creative. The topic naturally demands that we probe the ways in which we humans are similar to and differ from the artificial entities we are and will create.

    For many years, Doug Hofstadter gave talks during which he juxtaposed two recordings, one of a not particularly well-known musical piece by a composer such as Vivaldi, and the other a composition in the style of Vivaldi, but written by a computer program. He would then ask his audience to identify which was which. The exercise would be repeated with pieces by other composers, together with computer compositions that imitated the style of each of those composers. During a talk Doug gave at the Yale School of Music, he asked attendees to indicate on their ballots how much musical training they each had received. After listening to the samples, it wasn’t totally surprising that those of us with limited musical training were unable, for example, to distinguish the Vivaldi creation from the computer composition. However, what was fascinating was that the votes by the professional musicians and graduate students in music in the audience were only marginally more accurate than their less musically trained peers.

    Certainly the computer program that analyzed patterns and relationships in Vivaldi compositions, and then imitated these in producing a new composition, “at least appears to be creative, to some degree”, to use Maggie’s words. But is this creativity? That is the question we all want to ask. Of course the answer to that question has a lot to do with how one defines creativity or a creative act. However, this is not exactly the question Maggie is asking. She wants to explore whether computer models help us to understand human creativity?

    Maggie has spent perhaps as much time as anyone reflecting on the nature of creativity and the possibility of computers being creative. This is evident in the sensitivity and subtle use of language with which she develops her inquiry. The central question Maggie focuses upon is, however, the nuanced and cautious expression of someone well acquainted with the pitfalls of claiming that the kinds of computer technology we have today or will have in the near future are truly creative.

    I’ve spent the past few years fielding questions about whether computers can make moral decisions, and have also recently written an articles on what we can learn about human ethics from computers. I have come to appreciate that there is an art in how we enquire into the question of whether computers and robots might eventually emulate higher-order mental abilities including creativity, compassion, consciousness, self-awareness, and the ability to discern the desires, intentions, and goals of the humans with whom we interact. Maggie Boden has often demonstrated that she is a master of that art.

    Attempts to build any higher-order mental capacity into a machine demand that we think through the many processes that make up that activity with an unusually high degree of specificity. Implementing any facet of that mental capacity within a computer system or robot provides us with a model through which we can study whether we have or have not understood that aspect of intelligence. But as long as there are key elements of a process, such as being consciously engaged in being creative, that computer models fail to capture, there will be challenges to any claim that Picassobot is truly creative.

    Nevertheless, as Maggie notes, even experiments in building machines that engage in simple combinational creativity, drawing, or basic architecture can surprise us. Yes, the computers have been programmed by people, do not go through a conscious learning process where they explore a medium, fail to demonstrate a sophisticated capacity to evaluate and revise, and more importantly, can not develop their own criteria for evaluation. But oh my, their output can be unique, beautiful, and far beyond what I or other untrained practitioners can create.

    There is a fear that as we build more sophisticated computers human creativity will be pushed aside. Perhaps. But what is more likely is that we humans will explore new realms of expression. Many of these new forms of expression are and will emerge from the explorations we make using new media, including the near-ending stream of computer-based technologies.

  • Mark

    I would like to address a few points Margaret raises regarding BF-searches and the implementation of JAPE.

    Complete BF-searches are definitely implausible with current computing power. Partial BF-searches, however, are possible and often implemented. To illustrate these partial BF-searches, let me give an example from my personal experience. As a novice computer user attempting to figure out how to print from a spreadsheet application, for instance, I would conduct a BF-search of the application’s toolbars and menus until I found something related to printing. When using a word processor and attempting to print, I would again conduct a BF-search. Now with more experience (primarily heuristics and pattern recognition), I can quickly find the printing functionality of most applications without much searching. But I still conduct a search, even if I’m satisfied on the first attempt. Experience reduces my complete BF-search to a much-informed partial BF-search but a BF-search nonetheless.

    Admittedly, I think that purely I-creative acts occur only infrequently, and hence, BF-search techniques likely factor into most processes yielding creative output, even those processes of humans. I still think that computer creativity is much more BF-creative than I-creative given the necessarily strict and structured nature of computer programs. On the continuum from pure BF-creativity to pure I-creativity, I would place computers very nearly BF-creative.

    Margaret noted in her response that JAPE has a set of “structured templates” and “relevant criteria” it uses for creating jokes. Granted I am not familiar with JAPE’s software implementation, but I still find it hard to understand how JAPE generates jokes without conducting a partial BF-search. JAPE is given constraints, which describe the semantic or syntactical structure of jokes, and a “goodness” metric, which tells JAPE how to identify a good joke. I can think of no other way for JAPE to generate jokes other than generating combinations of words (or letters or phrases) that satisfy the constraint and then outputting those combinations that satisfy the “goodness” metric. JAPE may have methods for pruning its searches but the search methodology remains.

    I think the difficulty in implementing a program like JAPE without using the BF-search methodology lies in how to endow a computer program with I-creativity. A computer program is by definition a sequence of rules to be executed (ignoring the possibility of parallelization for now). The program can “break out” of its rules through self-modification, but then that capacity for self-modification must have been part of the rules at the outset — that’s not really “breaking out” of the rules. The only way I can see to sneak I-creativity into the program is through a source of randomness, genetic algorithms, or sufficiently complex neural networks. Margaret has already raised issues about the “creativity” of randomness, and since genetic algorithms and complex neural networks harness randomness in structured ways (mutations for genetic algorithms and training for neural networks), I’m not really sure that such techniques actually empower computers with I-creativity.

  • Margaret A. Boden


    Thanks for the many kind words!! With respect to your final paragraph, I entirely agree. Human creativity won ‘t be reduced. And, as you say, it’s already being increased by these technologies. For example, the various types of computer art (which are briefly classified/described in a few of the chapters in my forthcoming book CREATIVITY AND ART) offer a number of creative possibilities that were inacheivable, even unthinkable, just fifty years ago.

    More generally, there is a problem in discussions of this general area, because the word “computer” naturally makes people think about … COMPUTERS!! And, for most people, that means a PC or laptop, or something essentially similar, just bigger. There’s no question whatever, in my opinion, that human thought, language, and creativity could not be fully simulated by ANY current computer. Even if we could take the most powerful computers extant, and connect thousands of them together, that would still be true. The reason is partly that the human mind is s enormously rich, and so subtly powerful, in its contents and in its information-processing (e.g. associative memory) that although (I believe) it is **in principle** possible to match it in some artificial system, it’s not–and in my view never will be–possible **in practice**. — And the reason is also partly that we don’t understand enough about just what computational processes are involved. Moore’s Law is irrelevant here: the biggest problem is not our lack of processing power, but our failure (as yet) to understand the details of the processes and architectural structures that make up the virtual machine which is the human mind. We need to discover/define many new sorts of information-processing before we can get a good theoretical grip on these issues. —– It would have been better, in some ways, if I had spoken of “computer-based systems”, or even “information-processing systems”, instead of “computers”. However, my defence, here, is that part of what I wanted to do was to point to some actually-existing computer models of creativity and show how, in their failures as well in their successes, they help to cast light on **human** creativity.


    I think that some of our apparent disagreement, here, may be due to terminology. I don’t deny that “search” is typically involved in computer models of creativity, JAPE included. But heuristically-guided search of a space isn’t the same thing as exhaustive brute-force search, which is what Mark (in his first comment) was talking about.

    As for “I-creativity”, whatever this is, it isn’t magic. It’s an information-processing facility (or rather, a varied set of such) possessed by human minds which could, in principle, be expressed in computational terms–if only we knew how to do so. Many more types of computation are available to AI today than were available in its infancy. No doubt, further progress of this kind will be made. The fact that we can’t say today how a certain result could be achieved doesn’t mean that we’ll never be able to do so.

  • My own interest in the intersection between computers and creativity is, and has been since my undergraduate days, the potential for human-computer collaboration: using technology to extend or re-shape existing human capacities. I wrote a paper about it during my undergraduate exchange year at Sussex University, rather presumptuously called “Human-Computer Collaboration: The Vegar Effect”, after the Star Trek space probe that, by successfully “merging” with its human creator, is able to boldly go where neither artifact nor human could go before.

    The program at the heart of the paper was a murder mystery “generator” that was meant, together with its human operator, to write a short story. As with many undergraduate programming assignments, its ambitions far exceeded its abilities. Though my present doctoral studies are suitably more modest, nonetheless the original inspiration and the questions it raised for me remain: How, from words and concepts, do we structure our worlds, both experienced and fictional? How can computers help us understand that creative process – because it is deeply creative – and how can they best become part of that process? Where, ultimately, might human and computer go together that neither can go on its own?

    I should say something about what I mean by “creativity”, best summed up as recombining elements of past or present experience in strikingly novel ways that may yield insight or more immediately practical benefit. Creativity, like imagination, does not come out of nowhere. But though it comes forth from our presently lived world, at the same time it has the potential to tear down the walls of that world, to let them be built up again over some hopefully wider territory.

    Appropriately to the present discussion, much of my dissertation concerns conceptual spaces. My usage is based on Peter Gärdenfors’ conceptual spaces theory of concepts, which describes, as the subtitle of his 2004 book says, “the geometry of thought”. Concepts are, roughly speaking, structured units of thought, the building blocks of our ideas; but what sort of things are they? Rather than giving them a linguistic framing, as is all too often done, Gärdenfors describes concepts as convex shapes in multi-dimensional spaces, whose dimensions correspond to the integral properties of the concepts.

    Part of the intended contribution of my thesis is to describe how all of an individual’s many conceptual spaces, from Gärdenfors’ account, may be mapped together into a single unified space of spaces (something Gärdenfors sought to accomplish in his book but was not quite able to arrive at). Something similar, then, should happen at the societal level, where the many different conceptual spaces – one to each agent – get mapped into a unified space of the whole. Another mapping takes place to words of a language. Each level of mapping is imprecise and prone to error. One of the things the creative process does is exploit those ambiguities.

    Experience drives concept formation: the partitioning of our conceptual spaces. At the same time, concepts are, as I describe them, the expectations that drive experience, telling us what to look for. Concepts, which abstract away from the present moment, are constantly applied back to the moment – moment by moment. When confronted with an expectation breakdown, the conceptual agent has a choice of strategies:

    1. Adjust the structure of the closest matching concept until it fits, e.g. by adding new dimensions or re-arranging its components.

    2. Partition an as-yet-unpartitioned (unexplored) sector of the conceptual space.

    3.Most radically, remove the partitioning from some sector of the conceptual space and re-partition it. As opposed to conceptual change via 1 or 2, this is conceptual obsolescence and replacement.

    I am presently working on a toy application, which I hope to include in the last full chapter of my thesis, intended both as an illustration of the extensions I wish to make to Gärdenfors’ conceptual spaces and as a “proof of concept” for a next-generation mind-mapping tool: allowing users to create externalized models of some portion of their own conceptual spaces and (creatively) explore them. This is potentially quite valuable if you believe, as I do, as I think Maggie Boden does, that self-discovery is one of the primary founts of creativity; and, contra Star Trek, mind, not space, is the final frontier.

    Thanks to Maggie for a thought-provoking piece!

  • Ranulph Glanville

    Response to Maggie Boden
    This is a delightful introductory text, a delight to read and full of delicately presented points of importance. Like everything, it starts from its own particular presuppositions, and these are what I perhaps take issue with: in particular I take issue with attaching creativity to the parts of the system. I like the distinctions made between types of creativity, surprise and all the rest, but I don’t like the splitting of the experiential whole. So I’ll put a view of creativity that starts from a different place.

    Few, if any, systems that involve humans acting lack powerfully circular links. Feedback is perhaps the most familiar way we describe this, although feedback suggests a minor link back whereas, organisationally, the link back is as significant and important as the feed forward: that’s what circularity means. For example, when I paint, I hold a brush, dip it in a colour I select or create, and mark the canvas. Then I look at what I’ve done, and continue to paint (and rework), possibly after some long time. Eventually, I decide I’ve finished.

    Using painting to illustrate creativity might be considered a cliché. But I am interested in this understanding of creativity, that it arises in a system, not in the parts of a system. My criticism of Maggie’s approach is that it seems to be concentrated on the parts. I suppose we are all familiar with that other old cliché, “The whole is greater than the sum of its parts.” This is used to describe emergence, but it also tells us about a difference. Kenneth Boulding corrected the statement: “ The whole is different than the sum of its parts.” Difference being the key which has to do with identity: the system that is the whole is distinct and different from the sum of its parts.

    When I write at my wordprocessor, and I feel the rush of excitement of creativity, I don’t know where to place this creativity. Did it come from me? From the computer and its program? That is not how I experience it. I experience being in a creative moment with my computer, and sometimes I realise that at the time I experience nothing because I am lost in both the whole and the moment, united with my wordprocessor in the act of creative writing. I have no idea what comes from where, and I even consider this a meaningless distinction: what creativity there is comes from the two together. The creativity comes from the whole system, not from one of its components. It may, perhaps, be described as shared. The relationship between me and the machine is circular, any causality is circular, and the creativity is expressed in the links (the output from the computer is an input to the writer).

    Maggie’s account is not based in the experience of acting creatively. Typically, it attempts to stand outside and evaluate the parts of the creative whole, ignoring any evidence articulated from within the system itself: indeed, that’s one of the aims, to consider how the computer might help us to understand human creativity. In a sense, this is contradictory: to treat creativity this way is to ignore the critical element of being creative: it is an activity. The question thus becomes, How do we act creatively, with (or without) a computer? If I slide up Maggie’s scale, from creativity to intelligence, I would ask where is the intelligence when you and I meet. And my answer would be, in our behaviour together, not in me, or even you, but in us. I think creativity can be understood as occurring in this way, so that creative behaviour is found in (the links between the components of) the whole, rather than its parts. And when the observer is similarly linked in a circle, that means the creativity observed is the creativity of the observer and the observed (the “creative system”) together, not of the creative system by itself.

    So I assert that creativity might (should?) be understood as a property of wholes, not of parts in wholes. And it should be examined in the same manner, through the unity of the observed and the observer. This internalises the observer, undermining the carefully constructed delusion of the detached, objective observer, and that gives us the epistemological quandary which lies at the base of second order cybernetics, from which Francisco Varela learnt his “view from within”. But that’s the fun of it all.

    I believe Maggie implicitly recognises the dependency of this circularity, at least in part, in her concluding paragraph: …“Computers can come up with new ideas, and help people to do so.”…

  • Margaret Boden’s account offers a provocative framework for looking at creativity more generically, seeing how computers might throw light on the categories that are very usefully proposed. I’d be keen to debate whether these categories might open onto the kinds of creativity that are specifically born of and embedded in models of human computer interactions, i.e. that more incisively and recursively move betwixt and between humans and computers. This would relate the discussion to work undertaken by artists who are modeling richly interactive emergent systems, anchored in the participatory “social software” platforms developed by quite specific computer communities. When multi-user operations in distributed environments underpin workflow enactment models that elicit or solicit new kinds of encounters, creativity could be seen to occur at an excitingly, deeply transformational human computer juncture.

    Picking up on Alan Bundy’s rugby example, reflection on creativity and computer models could be extended by looking at the inventive synergies of teams. Rugby tricksters whose efforts challenge and transform rules – e.g. recently legalised lineout lifters – are engaged in the collectively, experientially driven creativity that characterises complex social-technical systems. Riding a loose analogy, one might say that creativity in so-called digital art involves models that are increasingly multi-agent, accommodative of and responsive to hybrid data sources (however far these must be reduced to become process-able elements). New forms of creativity are being modeled around platforms that prioritise input and feedback from individual and grouped human interactions, and from physical sensing systems that impart a “real-world” ethos to works which seems to potentiate their effects on existing conceptual systems. Consequently, means of encounter and hybridisation built into human computer interaction models are themselves powerful catalysts of transformational creativity. At the same time, they provide insights into our capacity as humans to develop supposedly impossible ideas, including creative notions of sociality. This approach perhaps links with Joel Parthemore’s and Ranulph Glanville’s postings in favour of systemic approaches to creativity and to constantly evolving human-computer feedback.

    Thanks Maggie for a wonderfully stimulating kick-start to a good game, where I’ve found it impossible not to refer to others’ enlightening inputs and feedback – this of course further emphasises my belief in the fact that creative processes in computers, humans, and in their inseparable mix, are eminently socially grounded!

  • The Wonder of Cogs

    Professor Boden’s rational analysis of creativity and conceptual space brings brilliant philosophical thinking to bear on creative practice. In my experience most artists are quick to agree that creativity isn’t a special ‘faculty’, though most would be equally likely to say that they experience creative insight as instinct rather than idea. How would an artist define instinct? – Often by describing a sensual response, telling a story or describing a particular occasion. The moment of origination is often experienced as feeling rather than thought – but for the artist, what follows? – Most likely the desire to juxtapose the original impulse with other, related activities rather than any desire to follow the moment of intuition with sequential thought.
    Grayson Perry tells the story, in his memoirs, of his father ‘bringing home a ball-bearing for me when he was working in Hoffman’s ball-bearing factory … It was heavy, shiny and perfect. My dad told me they made ball bearings by dropping blobs of molten metal through the air into cold water and, while in mid air, they formed into perfect spheres.’ I can see the shining metal spheres now, when I look at Perry’s art works – the shiny, silvery surfaces of his urns, beneath the overlaid inscriptions of graphic narratives. But it is I, the viewer, who am making that connection, the artist himself doesn’t make it. Instead, he proceeds to unmake it. The story continues, ‘My father’s shed was full of objects like ball-bearings : parts of engines, fixings, lengths of metal and cogs …’ Attention appears to wander as the response progresses from wonder and a vision of perfection or completion, straight back to the artist’s driving fascination for the unmade.
    Creavitity for the artist does not just create surprises; creativity is surprising; something ordinary needs to become unfamiliar, in a process the Russian Modernist poets called ‘making strange’. In artistic practice, hypothesis invariably signals paradox. Perry’s shiny surface is experienced as both ‘for me’ and alien; perfect and essentially incomprehensible – that is, wondrous, rather than explicable. In my examples, the desire to make a work of art seems contingent on, rather than a result of, conceptual thinking, originating somehow in a realm anterior to conceptual space – as Shelly implies in his description of the artist’s imagination, ‘which some invisible influence, like an inconstant wind, awakens to transitory brightness.’ The moment of origination is transitory; the drive that follows is towards making and completion. But perhaps that, too, can be rationalized …
    Dr. Sue Roe,
    May, 2010

  • Margaret A. Boden


    Your applications of Gardenfors’ conceptual spaces are interesting. But I think there’s a pun hidden here. As you say, he is concerned with **concepts**. And so are you. An d so am I, **when** i talk about combinational creativity. But combinational creativity is not the only type. Your definition of creativity as (only) recombnational fits well, I think, with a focus on concepts. But a set of generative stylistic constraints is not the same thing s a concept.

    I agree that computers can sometimes help us extend our own creativity, in ways that we couldn’t have managed without them. Some examples are seen in computer art (see the final chapters in my forthcoming book CREATIVITY AND ART). Other examples are seen in science–for instance, in visualizing/synthesizing new chemical molecules for pharmaceuticals, etc. (This is often done using evolutionary programming.)

    I’m suspicious of your sequence “Conceptual obsolescence and replacement”, because this doesn’t seem to take account of the **continuity** between the old concept/style and the new concept/style. Even the most radically transformational creativity has some continuity with previous thinking. Indeed, that’s what enables us to understand and appreciate it–although that understanding may take many years to develop, if the transformation is truly radical. (Lots of historical examples of this, in both science and art.)


    Yes, feedback is hugely important. It’s what’s going on in the valuation phase, at least when the creative person is doing their own valuing, and adjusting their work/ideas accordingly. And it’s what’s going on also in stigmergy—e.g. noticing the newly-uncovered veins in a piece of marble and altering the shape of the sculpture-in-process to take advantage of them. (Of course, one could also take steps to **avoid** the veins, but unless that leads to a significantly different creation, we don ‘t normally think of it as an example of stigmergy.) An excellent book on this topic is Andrew Harrison’s MAKING AND THINKING–published in 1978, I think.

    As for creativity being ascribed to the whole system, not to its parts, I agree–but the “system” can be thought of at the social level or at the level of the individual mind. I normally focus on the individual mind. Whether or not one takes explicit account of the social-cultural interactions, the functioning of the creative individual mind needs to be explained. Yes, it’s a dynamical feedbacjk system. But just what strictures, conventions, concepts, criteria ….. is it following and forming? I agree that one can find out very little about the answer to that question by introspection: as Ranulph says, the creator doesn’t consciously experience these things, or if s/he does, not in full detail. To try to explain the underlying processing (some of which may be open to consciousness, but much of which is not) isn’t the same thing as describing the conscious phenomenology of creative persons. (Which varies, anyway! And it varies **partly** because of their formal/informal **theories**, or **hunches**, about the nature of creativity. Introspection is just as theory-driven as perception of the external world.

    Ranulph, as a committed constructivist, won’t like my reference to “the external world” …. so I’d better shut up!!

  • Ranulph Glanville

    Dear Maggie,

    Thank you for your comment. I think that my point is that it is sometimes interesting to ask now what something is, but where/how we come across it.

    Just a tiny point of information:

    I fear that you, like so many, confuse constructivism with solipsism/idealism. Constructivists don’t, as far as I know, deny an external world (mind independent reality), but they don’t confirm it either. They just say that they cannot know, and that this not knowing should be taken seriously, not dismissed. They also say that, taking it seriously allows one to be a realist or an idealist, as one likes, but realising it’s a choice and a convenience, not a truth. This also means that the choice that is convenient can change.

    So I don’t object to the external world: only to the belief that we can know it objectively, as a truth!

    Constructivists also believe in the test of viability. Does my explanation hold? is the question.

    There are other points: I agree that the system, as I was pointing to it, may be at any level. That’s an easy one. The others might need a bit more work. Maybe I need to go back to more notes I made on your original (I really don’t want to get into a squabble). However, I’ve a busy night and day tomorrow, but hope to get back to you after that.


    By the way, why are people always so dismissive of solipsism?

  • Margaret A. Boden


    Thanks for reminding us of the richness of the new forms of social (distributed) creativity enabled by human-computer interactions. Yes, these are hugely exciting. The possibilities are literally infinite, because of the general-purpose nature of digital computers. The actual results will be limited by our own creative powers, of course, including the creativity of the software-writers as well as the participants. Already, many of these interactions are transformational, as you say. And I’m sure more will come.

    Do you foresee (or is it already here??) a time when some people act as professional performers/prompters, where they both contribute to or mould the emerging HC interactions and also prompt, or guide, or encourage the non-professional participants (a.k.a. the ‘audience’) to engage interactively in certain (hopefully, imaginative) ways?

    A bit like a jazz-gig where the improvising musicians encourage (not just invite) the audience to join in ….

    Has distributed HCI been exploited much, or at all, in experimental theatre?

    (Shades of Gordon Pask’s vision of a cybernetic theatre, at Joan Littlewood’s Round House?)


    Well, it seems to me that when artists say they experience creative insight as instinct rather than idea, they are saying what I referred to above as the phenomenon of intuition”. The word “instinct” is apt, here, insofar as it suggests the absence of conscious thought–and maybe also, in at least some cases, the driving motivation that one may experience while thinking/acting creatively.

    There’s a very good book on creativity by David Perkins, called THE MIND’S BEST WORK (Harvard U. Press, around 1980 I think), where he argues that the main difference between exceptionally creative people and the rest of us is not better cognitive skills on their part but more intense motivation.

    And another book, by Perkins’ colleague Howard Gardner, called CREATING MINDS (1993), seems to bear this out. He looks at seven very different twentieth-century creators (Freud, Stravinsky, Martha Graham, Einstein, Gandhi. T.S.Eliot, and Picasso), and asks what they were like as people. Not to put too fine a point on it, they were all utter b****rds”, horrendously selfish and virtually impossible to live with. But this was largely due to the fact that their ideas, in each case, were so very transformational that they were held to by crazy and/or offensive to virtually everyone around them. To have the self-confidence to persevere, nonetheless, and the will to do so despite many obstacles (both intrinsic to the task and socially generated by the group), requires a very special sort of personality–not a pleasant one to live with.

    I liked your example of Grayson Perry, and also your quote about “making strange”. Yes, human minds are so enormously rich, and our cultures so rich and fluid also, that there is infinite potential for the sorts of (largely idiosyncratic) associations you mention to arise—-especially where natural language is involved. That’s why I believe that there will never be a computer model of creativity that will match human creativity in the general case. And notice: that’s not saying “There will never be a computer Shakespeare”, but rather “There will never be a human-level Shakespeare-reader.” (“Human-level”, here, assumes a certain linguistic and cultural background, of course; but you and I are reasonably insightful Shakespeare-readers, without being Shakespeare!)

    Your quotation from Shelley is interesting. But, as you say, it’s description of the (phenomenology of the) creative experience. What interests me, also, is the explanation of that.


    Yes, some constructivists do indeed avoid denying thee existence of an independent external world. Ernst von Glasersfeld is a good example of this. But others dismiss this notion as an intellectually–and even politically– oppressive grand recit.

    However, there’s no knock-down argument either to prove or to disprove that sort of extreme (ontic, not just cognitive) constructivism.

    Solipsism: you may well ask! I don’t see how an ontic constructivist could avoid it….except that they would regard social/cultural discourse (sic) as an especially important/valuable type of discourse.


    We inhabit a richly articulated environment, with potential for astronomical combinatorics.

    It wasn’t thus for all our forerunners, such as microbes in chemical soups, where the only variations possible that could affect individual control decisions are changes in concentrations, concentration gradients, flow rates, and illumination, and perhaps amounts of neutral matter (e.g. sand particles) in the soup. The comparative simplicity, and paucity of information in such environments made it feasible for natural selection to produce pre-compiled answers to all the possible control problems that could arise.

    But many increases in complexity, in environments and in the organisms themselves, provided increased demands on the control mechanisms, which, as we’ll see, tended to shift the burden of decision making from the genome to capabilities and knowledge developed or acquired by the individual.

    Many different changes in the environment could increase the diversity of decision contexts for
    primeval soup-dwellers, including the presence of persistent structures on a relatively large scale — for instance a large rock with different required nutrients on different parts of the surface.

    For some feeders, randomly slithering or crawling around the surface and constantly grazing might suffice.

    However, if different nutrients are needed at different times, and the direction towards a required nutrient cannot be detected in local chemical gradients, and the food cannot be sensed at a distance, then our microbe needs the ability to take in and store information about how places are related, so that it can choose appropriate routes to the food it needs, when the need arises, starting in different locations.

    Some robots can do things like that — moving around and building re-usable maps of terrain or
    buildings (i.e. using SLAM mechanisms: for Simultaneous Localisation And Mapping).

    If the food is also mobile, and it attempts to escape being eaten by detecting an approaching grazer and moving away, or if there are competing grazers, then decisions about direction and speed of motion towards food become more complex. For instance if several competitors are
    already close to a source of the currently required nutrient then it may be better for a grazer to head for a more distant source of the required food, where there are no competitors.

    Of course, whereas previously the grazers could use local sensors and SLAM capabilities, they now need to be able sense what is happening at a distance and use that information when deciding which food source to head for. The greater the distance at which they can perceive things the more varied the combinations of information items on which to base decisions.

    If the food evolves a defensive shell so that a grazer can no longer simply approach and consume, but also needs to remove the shell, that would favour the evolution of beaks, teeth or claws that are capable of holding and breaking open the food.

    But now a hungry creature has more complex control decisions than merely selecting directions and speeds of movement to get to immediately consumable food: it also needs to control the motions of beak, teeth or claws in relation to individual food items. That can involve selecting a direction of approach if there are intervening hard obstacles and more detailed control of motion if the shell has recognisable weak points that need to be attacked. (Try searching the web for videos of parrots eating walnuts.)

    Such weapons that provide manipulation capabilities can also acquire other uses, e.g. damaging
    competitors for the same food, or removing obstacles that prevent access to the food.

    In many situations, this will increase the variety of options for action available to individuals and the variety of sequences in which actions are performed. E.g. in which order should one (a) move towards food, (b) look for competitors for the food, (c) move toward competitors, (d) threaten or attack competitors, (e) start eating food, (f) decide whether to retreat, …. and more.

    Evolution has at least the options of either producing innate control systems that produce behaviours that successfully achieve consumption of food, or developing some sort of learning mechanism that enables the best control decisions to be associated with various sensed conditions to be learnt by each individual. The former is feasible if conditions remain static for a long time, so that relatively fixed (though not necessarily simple) designs can be evolved. If the world changes too fast, evolution will not be able to catch up, and species whose individual members can learn will have an advantage.

    There are many more scenarios that can be imagined, which are not necessarily entirely fanciful
    — in view of the enormous diversity of life forms we find on earth. Some scenarios involve the food evolving to live on dry land out of reach of its aquatic consumers. If the aquatic food source runs out then evolution could favour variants of the grazers that develop the ability to move onto dry land to meet their needs.

    If the food organisms then develop stalks supporting the edible portions out of reach of the grazers, the result could be evolution of jumping, or climbing abilities, or long necks, or long limbs supporting the manipulators, or abilities to build climbing supports or to throw things to bring down food.

    All changes in morphology of the food-seekers are worthless without corresponding changes in
    control mechanisms — requiring more and more varied kinds of information in more and more
    complex situations to be used in acquiring food.

    Some of the foods may develop camouflage, by evolving towards an external appearance that
    resembles that of other inedible objects. In that case the feeders may have to develop more
    sophisticated perceptual abilities, perhaps taking account not only of the appearance of the edible items but also their location and perhaps how they respond to various tests, such as the sound produced when they are tapped with something hard. Alternatively, instead of directly evolving those abilities, evolution can produce mechanisms that learn in an individual’s lifetime how to do it.

    However, if individuals are not born or hatched with all the competences they will need, then they will take some time to learn, during which they require help from adults — protection,
    feeding, and opportunities to learn. This will add to the information processing demands on the
    adults, who then need to detect and cope with not only the environmental situations that are relevant to potential harm and benefit to themselves but also those that can affect their offspring. They need to be able to detect vicarious affordances. This adds to the learning requirements of the offspring, if they have to learn to look after their offspring.

    The point of all this is that as the environment poses new challenges, evolution can produce responses that involve various combinations of change of behaviour using the original morphology, and change of morphology allowing new behaviours and requiring new control mechanisms, which can indirectly lead to yet more changes requiring even more complex behaviours and more complex control systems.

    The idea of an “evolutionary arms race” is not new. My examples are presented only in order to
    emphasise the increasing demands on control mechanisms and the need for increasingly sophisticated information-processing systems involved in such control.

    These are not easy for biologists to study because internal information processing mechanisms and behaviours and not accessible to normal methods of observation and measurement. They also do not leave fossil records.

    There is much more that can be said about the changing requirements for (a) forms of representation in which information is acquired, stored manipulated, interpreted and used; (b) the changing ontologies required as more complex types of information about states and processes in the environment are acquired and used; (c) the changing architectures in which more and more varied information-processing competences are combined, a process that itself requires more control decisions e.g. about which competences to activate and how to deal with conflicting control decisions generated internally; (d) the changing trade-offs between evolutionary changes in the genome and the processes of learning and development, partly under the control of the environment, in individuals.

    I return to my starting observation: We inhabit a richly articulated environment, with potential for astronomical combinatorics. When the wind blows there are many possible combinations of strength, direction, temperature, humidity, whether the motion is linear or curved (as in tornadoes), and many different kinds of airborne objects possible, e.g. sand, leaves, twigs, discarded rubbish, smoke, pollen, fumes from chemical plants, etc.

    When you stand in a forest surrounded by trees, bushes and other objects there are many
    possible directions in which you can move which will vary your relationships to other things in
    significantly different ways: bumping into a tree-trunk, bumping into a blade of grass and bumping into a wasp nest will have very different effects.

    Even if you remain still and do nothing, other people, other animals and things moved by wind or water can produce changing relationships that you may need to take into account.

    For an animal with mobile anipulators there are even more possibilities that can be combined in different ways with the possible changes and movements in the environment.

    Some numbers

    If there are a hundred different things in your environment each of which can change or remain
    as it is, then the possible number of distinct combinations is 1267650600228229401496703205376.
    (i.e. 2 to the power 100).

    If your learning procedures allow you to experience each of those possible combinations, and you try one every second, then the time required would be approximately 40196900000000000000000 years (dividing the number of combinations by the number of seconds in a year).

    If each of the changeable items has more than two options the numbers will be even larger. If we consider not just instantaneous changes, but sequences of changes over a period of a few seconds the numbers of possible sequences of changes will be even larger.

    An implication of all this is that neither evolutionary time scales, nor individual learning times available in the life of an individual can cope with exhaustive learning.

    Some reduction in the number of learning episodes required can be achieved by ‘chunking’, i.e.
    grouping together things that do not differ very much. But even that will not tame the combinatorics, as has long been evident in the case of language (as Chomsky, among others, pointed out).

    I am completely certain that you have never previously encountered the sequence of words in the
    paragraph you are now reading, unless you deliberately reread it. But that novelty does not stop you understanding what I am saying. You have ways of WORKING OUT the meaning, on the basis of your understanding of individual words and phrases and the ways they have been put together.

    Similar things can be said about many actions a child performs: some of them are trial and error behaviours, or copying behaviours, but many are novel solutions to problems (novel as far as the child is concerned) where the child is able to work out what will happen by combining previously acquired knowledge and reasoning about it — for example when a child works out for the first time that the edge of a rigid circular object can be used as a screwdriver, or who invents a new practical joke to play on a sibling, or who first devises an efficient strategy for nesting cups, after finding that randomly choosing a ‘next’ cup leads to frequent wasteful back-tracking.

    One such strategy is to start with the largest cup, then always seek the largest remaining cup to insert next. You probably worked that out for yourself long ago, without even being aware that you had done so.

    All this is a brief introduction to the study of the many ways in which biological evolution was under pressure to provided humans and other animals with information-processing mechanisms that are capable of acquiring many different kinds of information and then developing novel ways of using that information to solve any of millions of different problems without having to learn solutions by trial and error, without having to be taught, and without having to imitate behaviour of others. I.e. they are P-creative solutions.

    I conjecture that these highly practical forms of creativity, which are obviously important in crafts, engineering, science, and many everyday activities at home or at work, are closely related to the mechanisms that also produce artistic forms of creativity. But for that we have to go into the complex topic of where motives come from, and how alternatives are evaluated, about which I have said nothing so far.

    One of the problems of AI researchers is that too often they start off with an inadequate
    understanding of the problems and believe that solutions are only a few years away. We need an educational system that not only teaches techniques and solutions, but also an understanding of problems and their difficulty — which can come from a broader multi-disciplinary education. That could speed up progress. It might even be a creative solution to an educational problem.

    Of course, none of this will impress people who don’t WANT to believe that machines can be creative. They just need to learn to think more creatively.

    I have more on these topics in presentations here.

    • I forgot to make it clear, though I hope it was obvious, that my response is merely an elaboration of some of the points Maggie had already made — emphasising the importance of studying cognition and creativity not merely in the context of humans and future machines, but also in other products of biological evolution, taking full account of the features of the environment that helped to define the problems evolution solved, some of them rather unobvious problems.

  • Ah, thank you Maggie for this prompt. Some exciting art pieces have engaged audiences as contributors to and moulders of emerging HC interactions – I remain attached to pioneering work like Anonymous Muttering by Knowbotic Research, which goes back to the mid-nineties and generated a powerful light and sound installation by throwing very different kinds of inputs into a massive processing “crucible”, including those of spectators manipulating various sensor arrays. Because it was sensorially quite overwhelming, it freed us up to enjoy (or not!) complex emergent processes that thoroughly outstripped (the then prevalent) imperatives of immediate legibility and causality that were constraining (too) much interactive art.

    In a more specifically theatrical register, perhaps closer to the Pask-Littlewood Fun Palace orientation, groups like Konic Theatre in Barcelona ( have been working for the last decade on facilitating audience participation with interactive devices, and individuals like Chris Salter (, notably with projects like TGarden and Chronopolis, have been developing “theatre” based or related work to see how people can generate aesthetically meaningful events from within responsive media environments. There’s a lot going on – and much of it in my opinion can be pretty tedious if it’s too caught up in tool development processes; sometimes however the creative miracle happens and the mix of affordances, human and technical (openness and playfulness first and foremost, though of course constraints are flipside affordances in the creative world), and that serendipitous thing called timing, patterns of encounter, whatever (Victor Turner would talk about “intersubjective illumination”) generates something meaningful and memorable.

    In short, YES! There has been and is exploration of distributed HCI in theatre or something akin to it, and it can – rarely of course – produce the gleam of deeply creative endeavour.

  • Margaret A. Boden


    Yes, P-creativity (as well as learning) is going to be needed if the environmental/behavioural combinatorics are astronomical. As for crearivitty in arts vs. science, it seems to me that the same broad principles of creativity are at work in both cases, with respect to the **origination** of new ideas–although of course there will be detailed differences from cases to case. But the **evaluation** pf the new idea is very different indeed in the two cases. E.g. an artwork doesn’t necessarily have to ‘match’ the external world in any detail (although many do: think of fifteenth.sixteenth century Dutch interiors and still lifes).

    You’re right: motivation is hugely important.

    There’s very good reason to believe that creativity is crucial in sexual selection: see the FASCINATING book by Geoffrey Miller, THE MATING MIND. (If anyone picks it up and finds the first few chapters–on comparative anatomy and on palaeontology– dull and/or difficult, just skip them. They are in fact highly relevant to Geoffrey’s main argument, but they’re not essential and may put some readers off.

  • Thanks for the stimulating article Margaret!

    I’ll like to delve a bit on evolutionary considerations. Aaron has hinted at how evolutionary processes and adaptation to complex environmental needs drive creativity/creative solutions in organic evolution. I would like to go in slightly different direction- that of considering the creative process as a Darwinian process in itself.

    I presume you would be familiar with Campbell’s Blind Variation and Selective Retention (BVSR) model to explain the creative process. In short its basic premise is that ideas are randomly combined (maybe unconsciously) and a selective value criterion applied that then selects a few while discarding others (and maybe only the finals selected few rise to consciousness).

    This I think fits nicely with your notions of combinatorial creativity and surprise being one factor of creativity; while Being valuable being another criterion. I wish to expand the model further and draw a few more analogies.

    Creative process may be broken into two parts- the novel idea/artifact generation process and the subsequent/parallel evaluation process that evaluates the idea/artifact for usefulness. The second process may run at society level or in an individuals head as presumably only novel (unconscious ) ideas that fit the individuals aesthetic/intellectual criterion will be allowed to surface to consciousness. Here it is pertinent to note that access to consciousness is a scarce resource and this puts selection process on the umpteen novel combinations/transformations/explorations that are unconsciously produced to be of value and having adaptive significance.

    The first part of creative process that of novel idea/artifact generation can be broken into three parts- combinatorial creativity (combining familiar ideas to produce novel combinations)- this is like recombination of genes by sexual mating in subsequent offspring in organic evolution; exploratory creativity (exploring a conceptual space)- I see the conceptual space as nothing but ‘memes’ making up a particular conceptual space and the conceptual space (classical style of music for example)is defined by the frequency of ‘memes in that meme pool’. Now recombining the memes is already one way to achieve creativity/adaptation. However just like mutation of genes and subsequent selection/genetic drift changes the frequency of genes in the genetic pool of a species/population, so does the mutations in memes change the conceptual space over time. This is akin to exploring the possible space of that conceptual space to get the best possible fit with the selection criterion of that environment/conceptual space. The third type of creativity- transformational (transforming a given conceptual pace to a new one- having the impossible idea)- is like speciation as per my model. When the meme pool becomes such that two different styles are co-existent in that meme pool such that interactions or recombinations between them become less and less probable- we have the arrival of a new conceptual space or the splitting of original conceptual space into two.

    The second part of the creative process is selecting the right ideas/artifacts. Here again two poles of intellectual and aesthetic criterion can be applied. While the former merits an adherence to the value of TRUTH and coherence with external world; the latter depends on an adherence to the value of BEAUTY and internal self coherence. Of course truth and beauty are not the only criterion for selecting ideas/artifacts but they represent prototype criterion.

    Of course I approach these topics from varied other angles like the relationship of this creative adaptivity to personality traits/emotions, but will love to hear Boden’s Views on this from a purely creativity perspective.

  • As usual, Maggie has given a wonderfully concise and clear defence of a useful philosophical position. I’d like to take the opportunity to add a couple of points which are relevant to the discussion, but which have not been made explicit, in the context of human creativity and models thereof.

    Alan Bundy gives Maggie’s three types of creativity numbers, and I’ll use these below. In my own 2001 formalisation of Maggie’s theory (which refers to Alan’s 1994 publication) I show precisely how a creative system, conceptualised in these terms, can flip from the level of types 1 and 2 to type 3, through the device of reflection – essentially, representing itself. It also follows from the formalism that types 1 and 2 are instances of the same phenomenon: exploration of the conceptual space. One interesting question that this view leads me to ask is: what happens when the envelope of the conceptual space gets pushed, at the exploratory level? I’ve proposed several particular cases, and suggested how aberration (meaning exploratory creative effort that goes beyond expectation in some way) can drive creativity at the transformational level. (Wiggins, 2006a,b)

    The musical example is a good one here. From Mozart through to Wagner and Strauß, Western tonal music became (more or less monotonically) more chromatic, until, at the end, its very tonality began to fray. This, although it was incremental, was definitely transformation creativity (and there are perceptual reasons why it works in the way it does – not enough time to discuss all that here). However, it is not transformational creativity that sets out to be a paradigm shift, because it is expressed in terms of the existing conceptual space of tonal music. Schoenberg’s altogether more radical step, of “completing” chromaticism by throwing out the tonality, is considered, deliberate transformational creativity, specifically aimed at overthrowing the rules of the formerly accepted conceptual space – a very modernist approach, which could probably only happen in the 20th century. I cover this example in more detail in my 2003 paper (Wiggins, 2003).

    Formally, then, this is a fairly simple system: there is a search mechanism, and there is a specification of the space being searched. However, that system is only good for modelling an individual action alone in a fixed cultural space. Ultimately, any theory needs to take into account multiple agency, shared and private norms, and so on. That’s future work for me.

    Nevertheless, the simple system allows us to focus on one very important aspect which is implicit in the discussion above: the role of consciousness. I’m quite sure that all the participants in the discussion have a position on the role of consciousness in creativity, but no-one has stated it explicitly. I think that in order to answer Mark’s point, contrasting BF-creativity and I-creativity, we need to bring consciousness into our model.

    When we become aware of having an idea, we see it as more or less creative, and that is evidently part of Maggie’s “evaluation”. But, the way I see it, there is more to evaluation than this: in the part of the process of which we are NOT aware. One model of the process of inspiration is that an idea does not simply appear in the mind, but instead is constructed non-consciously, prior to the drawing of conscious attention to its existence. There is neurophysiological evidence that this is the case with minor everyday creativity in language generation: linguistic centres in the brain begin working well before subjective reports of awareness of the fact (e.g., Carota et al, 2009). Therefore, I-creativity may well be the upshot of BF-creativity at the non-conscious level, coupled with an equally non-conscious evaluation mechanism, which is capable of drawing conscious attention when a sufficiently good idea is produced. This point throws the BF- vs I- distinction into question, because the same mechanism can explain both, depending on where one places the veil between conscious awareness and non-conscious processing.

    (Incidentally, I don’t think the Edison example really works as an exampled of BF-creativity. The creativity involved here was what produced the notion of the incandescent filament, and that was imagined or serendipitous – I’m not sure if Edison recorded which. The brute force part is a kind of optimisation strategy, trying to meet the specification defined by the prior creative act.)

    In short, the models we’re currently working are only really scratching the surface, and we need to think beyond them, imagining how they fit into a larger picture, even though we can’t model that picture yet. However, I’m quite convinced that Maggie’s approach is a useful one, and I look forward, in future, to extending my own more mechanistic version further in some of the directions discussed above.


    Carota, F., Posada, A., Harquel, S., Delpuech, C., Bertrand, O., and Sirigu, A. (2009). Neural Dynamics of the Intention to Speak. Cereb. Cortex, page bhp255.

    Wiggins, G. A. (2003). Categorising creative systems. In Bento, C., Cardoso, A., and Gero, J., editors, Proceedings of the IJCAI’03 workshop on Creative S ystems. IJCAI.

    Wiggins, G. A. (2006a). A preliminary framework for description, analysis and comparison of creative systems. Journal of K nowledge Based S ystems, 19(7):449–458.

    Wiggins, G. A. (2006b). Searching for computational creativity. New Generation Computing, 24(3):209–222.