The Concept of Machine Creativity by Margaret Boden
Autorin: Janina Bodendörfer
- Introduction
The Concept of Machine Creativity, coined by British cognitive scientist Margaret Boden, is a theory identifying three different types of human creativity and aiming to transfer them to machines. Similar to Alan Turing, who replaced the question „Can a computer be intelligent?“ with a test (called the Imitation Game), Boden replaces the question „Can a computer be creative?“ with her concept of machine creativity. While Turing’s test is supposed to show whether a computer can give the impression of being intelligent, analogously, Boden’s concept should help to determine whether a computer can give the impression of being creative.[1]
- What is creativity?
According to Boden, an idea can be called creative, if it is „new, surprising, and valuable“[2]. Despite being aware of the inherent ambiguity in every word that this definition is made of, she tries to explain them as detailed as possible. An idea can refer to a specific concept the artist wants to express through their art. However, it can also refer to a „method for producing artefacts (a new type of paintbrush, a revolutionary camera or developing technique, or a novel way of casting bronze)“[3] , or a certain style for painting, sculpting, writing etc.
Boden differentiates between two different meanings of the word new, depending on the degree of novelty. The idea can either simply be new to the single person who came up with it, which means that it is a „first-time occurrence within their particular mental biography“[4]. She calls this instance P-creativity, with P standing for both „person“ as well as „psychological“. On the other hand, the idea can be new in the history of humankind, if nobody else has ever thought of it so far. Boden names this type H-creativity. Naturally, every idea considered to be an instance of H-creativity is at the same time an instance of P-creativity, which is why the latter is the concept that is more important, according to Boden. [5]
For her second criterion, surprise, Boden offers three different meanings, depending on the reason that has evoked the reaction of surprise. One can be surprised because of the statistical unusualness and therefore unexpectedness of something that has occurred. The second reason is that the individual having that reaction previously has not been aware of the possibility of the idea, and the third reason goes even further, in the sense that the idea has been considered as something impossible and counterintuitive.
Lastly, Boden states that the criterion of value has many meanings and is the one that is impossible to pin down. Its definition depends on various things, like the specific creative area or genre or the historical context, and those are subject to potentially rapid and unexpected changes, which is why it is very hard to define what makes a piece of art „valuable“.
- Three categories of creativity
Boden’s three categories of creativity, combinational, exploratory, and transformational are „distinguished by the types of psychological process that are involved in generating the new idea“[6].
3.1 Combinational Creativity
The first category is called combinational creativity. As the name suggests, it involves the creation of something unfamiliar by combining familiar ideas in a new way, much like the way master chefs have combined culinary ideas from all around the world or like a visual artist mixing completely different constructs together in order to create something new. [7]
This kind of creativity is something that can be mastered rather easily by an AI, as it simply involves a random recombination of existing ideas. To name only one example, there is the computer program JAPE, developed in 1993 by Ki Binsted at the University of Edinburgh, which generates punning riddles.[8]
3.2 Exploratory Creativity
The second category is called exploratory creativity. It is characterized by the exploration of the hidden potential that has not yet been discovered or sufficiently exploited within a specific conceptual space. By exploring what is already there, the artist is „extending the limits of what is possible while remaining bound by the rules“[9]. According to Boden’s belief, this type of creativity constitutes 97 per cent of human creativity. One human example would be the music of Bach, whose preludes and fugues stand at the end of a long journey, an exploration of tonality, untertaken by Baroque composers.[10]
A computer, with the ability to perform calculations a lot faster than the human brain, excels at this kind of creativity, by pushing the boundaries of a specific set of rules or patterns. In the 1980s, David Cope developed a computer program for musical compositions, called EMI/EMMY (Experiments in Musical Intelligence). Its task is to analyze and deconstruct a composition, identify the style, and create new patterns from the analyzed components. [11]
3.3 Transformational Creativity
The third and final category is called transformational creativity. Its defining characteristic is the transcendence of boundaries of the previously given conceptual space. Thereby, a new structure of thought is developed that makes it possible to create art that would not have been possible to create before. In most cases, these transformations are based on the change or abolition of certain rules that were essential before. One human example can be discovered in the transition from the musical epoch of Classicism to Romaticism, where composers like Schubert made a conscious decision to break expectations in a completely new way that was previously unimaginable.[12]
According to Boden, this type of creativity has been quite hard to master for computers. However, Du Sautoy argues that there is reason to believe that the rise of deep learning algorithms and artificial neural networks may have brought us closer to achieving this.[13]
- Criticism
In my opinion, there are several things to criticize about Boden’s concept of machine creativity. First and foremost, the biggest problem is for sure the fuzziness of the terms on which she bases her entire theory. The three criteria for creativity: novelty, surprise and value are certainly things to consider when it comes to the concept of creativity, but to use them as the sole pillars of the concept seems rather arbitrary and simply not satisfactory, especially considering that these terms themselves are extremely difficult to define and highly dependent on (subjective) human judgement. It follows from this criticism that her categories of creativity are similarly unconvincing, especially the category of combinational creativity. Can a machine that randomly assembles verses of existing poems into a new poem really be said to have performed a creative act? I would answer this question with no, because it would mean that all it takes to be creative is the element of randomness.
In conclusion, I think that Boden’s concept of machine creativity is possibly an interesting starting point, but it only shows that we first need to clarify and specify our concepts of creativity and art, before we can talk about how machine-generated art fits into this and what it really means for a computer to be creative.
[1] cf. Sudmann, Andreas. Computerkreativität. Maschinelles Lernen und die Künste Künstlicher Intelligenzen. In: Bernhard J. Dotzler/Berkan Karpat (Eds.), Götzendämmerung – Kunst und Künstliche Intelligenz, Bielefeld: transcript Verlag, 2021, p. 85-98.
[2] Boden, Margaret. Creativity and art : Three roads to surprise, Oxford University Press, Incorporated, 2010, p. 70.
[3] Boden, 2010, p. 70.
[4] Boden, 2010, p. 71.
[5] cf. Boden, 2010, p. 71.
[6] Boden, 2010, p. 72.
[7] cf. Du Sautoy, Marcus. The Creativity Code: How AI is learning to write, paint and think, London: 4th Estate, 2019, p. 11.
[8] cf. Sudmann, 2021, p. 92.
[9] Du Sautoy, 2019, pp. 9-10.
[10] cf. Du Sautoy, 2019, pp. 9-10.
[11] cf. Sudmann, 2021, p. 93.
[12] cf. Du Sautoy, 2019, p. 12.
[13] cf. Du Sautoy, 2019, p. 12.