6.9 KiB
| title | chunk | source | category | tags | date_saved | instance |
|---|---|---|---|---|---|---|
| Creativity | 6/14 | https://en.wikipedia.org/wiki/Creativity | reference | science, encyclopedia | 2026-05-05T07:11:10.173371+00:00 | kb-cron |
Jürgen Schmidhuber's formal theory of creativity postulates that creativity, curiosity, and interestingness are by-products of a simple computational principle for measuring and optimizing learning progress. Consider an agent able to manipulate its environment and thus its own sensory inputs. The agent can use a black box optimization method such as reinforcement learning to learn, through informed trial and error, sequences of actions that maximize the expected sum of its future reward signals. There are extrinsic reward signals for achieving externally given goals, such as finding food when hungry. But for Schmidhuber's objective function to be maximized also includes an additional, intrinsic term to model "wow-effects". This non-standard term motivates purely creative behavior of the agent, even when there are no external goals. A wow-effect is formally defined as follows: as the agent is creating and predicting and encoding the continually growing history of actions and sensory inputs, it keeps improving the predictor or encoder, which can be implemented as an artificial neural network, or some other machine learning device, that can exploit regularities in the data to improve its performance over time. The improvements can be measured precisely, by computing the difference in computational costs (storage size, number of required synapses, errors, time) needed to encode new observations before and after learning. This difference depends on the encoder's present subjective knowledge, which changes over time, but the theory formally takes this into account. The cost difference measures the strength of the present wow-effect due to sudden improvements in data compression or computational speed. It becomes an intrinsic reward signal for the action selector. The objective function thus motivates the action optimizer to create action sequences that cause more wow-effects. Irregular, random data (or noise) do not permit any wow-effects or learning progress, and thus are "boring" by nature (providing no reward). Already-known and predictable regularities also are boring. Temporarily interesting are only the initially unknown, novel, regular patterns in both actions and observations. This motivates the agent to perform continual, open-ended, active, creative exploration. Schmidhuber's work is highly influential in intrinsic motivation, which has emerged as a research topic in the study of artificial intelligence and robotics. According to Schmidhuber, his objective function explains the activities of scientists, artists, and comedians. For example, physicists are motivated to create experiments leading to observations that obey previously unpublished physical laws, permitting better data compression. Likewise, composers receive intrinsic reward for creating non-arbitrary melodies with unexpected but regular harmonies that permit wow-effects through data compression improvements. Similarly, a comedian gets an intrinsic reward for "inventing a novel joke with an unexpected punch line, related to the beginning of the story in an initially unexpected but quickly learnable way that also allows for better compression of the perceived data." Schmidhuber augured that computer hardware advances would greatly scale up rudimentary artificial scientists and artists. He used the theory to create low-complexity art and an attractive human face.
== Personal assessment ==
=== Psychometric approaches ===
==== History ==== J. P. Guilford's group, which pioneered the modern psychometric study of creativity, constructed several performance-based tests to measure creativity in 1967, including asking participants to write original titles for a story with a given plot, asking participants to come up with unusual uses for everyday objects such as bricks, and asking participants to generate a list of consequences of unexpected events, such as the loss of gravity. Guilford was trying to create a model for intellect as a whole, but in doing so, he also created a model for creativity. Guilford assumed that creativity was not an abstract concept, which was an important assumption needed for creativity research. The idea that creativity was a category, rather than a single concept, enabled other researchers to look at creativity from a new perspective. Additionally, Guilford hypothesized one of the first models that specified the components of creativity. He explained that creativity was a result of having three qualities: the ability to recognize problems, "fluency", and "flexibility". "Fluency" encompassed "ideational fluency", or the ability to rapidly produce a variety of ideas fulfilling stated requirements; "associational fluency", or the ability to generate a list of words associated with a given word; and "expressional fluency", or the ability to organize words into larger units such as phrases, sentences, and paragraphs. "Flexibility" encompassed both "spontaneous flexibility", or the general ability to be flexible, and "adaptive flexibility", or the ability to produces responses that are novel and of high quality. This represents the base model which several researchers would alter to produce their own new theories of creativity years later. Building on Guilford's work, tests were developed, sometimes called "divergent thinking" (DT) tests, which have been both praised and criticized. One example is the Torrance Tests of Creative Thinking developed in 1966. These test set forth tasks requiring divergent thinking, as well as other problem-solving skills, the tests being scored according to four categories: "fluency", the total number of meaningful, and relevant, ideas generated; "flexibility", the number of different categories of responses; "originality", the statistical rarity of the responses; and "elaboration", the amount of detail given.
==== Computer scoring ==== Considerable progress has been made in the automated scoring of divergent-thinking tests, using a semantic approach. When compared to human raters, natural language processing (NLP) techniques are reliable and valid for the scoring of originality. Computer programs were able to achieve a correlation to human graders of 0.60 and 0.72. Semantic networks also devise originality scores that yield significant correlations with socio-personal measures. A team of researchers led by James C. Kaufman and Mark A. Runco combined expertise in creativity research, natural language processing, computational linguistics, and statistical data analysis to devise a scalable system for computerized automated testing: the SparcIt Creativity Index Testing system. This system enabled automated scoring of DT tests that is reliable, objective, and scalable, thus addressing most of the issues of DT tests that had been found and reported. The resultant computer system was able to achieve a correlation to human graders of 0.73.