Journal of Creativity 33 (2023) 100063

Contents lists available at ScienceDirect

Journal of Creativity
journal homepage: [Link]/locate/yjoc

AI can only produce artificial creativity

Mark A. Runco 1
Southern Oregon University, USA

A R T I C L E I N F O A B S T R A C T

Keywords: This article (a) draws from various theories of creativity (e.g., 4P and 6P theories) and (b) uses several concepts
Artificial creativity from the creativity literature (e.g., self-actualization, emergence) to evaluate the claim that AI can be creative.
AI This approach suggests that, at most, the output of AI represents products which, although lacking, may be
authenticity
attributed with creativity. Such attributions are often mistaken, and, significantly, products say little about the
Intentionality
Pseudo-creativity
underlying process. Indeed, criticisms previously leveled at the view that the social recognition of products is
Standard definition of creativity required of creativity also apply to AI output. Several examples of products and overt actions that have been
Emergence mistakenly attributed with creativity are discussed. The most telling of these is the ostensible emergence by a
machine. The conclusion is that it makes no sense to refer to “creative AI.” One alternative is to extend the
concept of “artificial intelligence” to creativity, which gives us “artificial creativity” as the label for what
computers can do. Artificial creativity may be original and effective but it lacks several things that characterize
human creativity. Thus it may be the most accurate to recognize that the output of AI as a kind of pseudo-
creativity.

AI has become one of the most compelling topics in the social and more creative than 95% of humans.” Since attributions are subjective (as
computer sciences. The possibility of creative AI is often part of the are all judgments), they may be quite wrong but misleading. Yet certain
discussion. That is in part due to the fact that the possibility of creativity audiences may accept them and believe that AI is creative.
gets right at central questions of AI being independent of human input. It Existing theories of creativity support the conclusion that AI is not
is also relevant to the question of what kind of work AI can assume. Then and cannot be authentically creative. Admittedly, the current standard
there is the view that, if AI can be creative, various threats to humanity definition (Runco & Jaeger, 2012) is not comprehensive enough to reject
are possible (cf. Edwards, 2023). This article draws from creativity the pseudo-creativity of AI. This leaves us with two options: either we
research and theory to explore the position that AI is indeed creative. accept that AI is creative (because it satisfies the existing standard
Come to find out, a case can be made that AI will never be creative. definition) or we update the standard definition so it distinguishes the
Admittedly, in some ways it may not matter whether or not AI can ever authentic creativity of humans from the artificial creativity of AI. I
truly be creative. Businesses are strongly in favor of AI and will treat it as already proposed updates to the standard definition of creativity
if it is creative even if it is not authentically so. Google’s chatbot Bard (Runco, 2023a). That proposal identified two criteria and key di­
already states, “I’m Bard, your creative and helpful collaborator.” One mensions of creativity (authenticity and intentionality) which should be
virtual influencer on social media offers “boundless creativity” (quoted added to those currently in the standard definition (i.e., originality and
from [Link] see also [Link] effectiveness). The current article (based on two presentations, Runco,
com/milla-sophia-ai/). Also problematic is that claims of creative AI 2023c, 2023d) has a different emphasis. It suggests that the process used
focus on output (“products” in the vernacular of the creativity research) by humans when they are creative is impossible for AI. This is very
rather than processes, and these products may fool some people into important because theories describing processes (or what Jay and Per­
believing in creative AI. More precisely, some people may attribute kins [1997] called mechanisms) have much more explanatory power
creativity to original or surprising products of AI. Consider Plain’s than theories that recognize only outcomes, output, and products.
(2023) claim, “Artists beware: ‘Game changer’ test results show AI is This article begins with the question, What is necessary for creativity

E-mail address: runcom@[Link].

1
Ning Hao was the Action Editor for the review of this article. It is based on an address (“AI Cannot be Creative, but it may not Matter”) given when the author
received the Lifetime Achievement Award from Southern Oregon University, May 2023. The author appreciates the input of Jeffrey Tsao, Shulamith Kreitler, and
Robert Edgell. [Link]

[Link]
Received 7 August 2023; Accepted 7 August 2023
Available online 25 August 2023
2713-3745/© 2023 The Author. Published by Elsevier Ltd on behalf of Academy of Creativity. This is an open access article under the CC BY-NC-ND license
([Link]
M.A. Runco Journal of Creativity 33 (2023) 100063

but lacking in AI? It then contrasts products (from AI and elsewhere) creativity (Albert, 1996; Amabile, 1990; Runco et al., 1999; Sternberg,
with the processes used by humans when they create. The discussion, 2000) but lacking in AI. AI does not choose to produce novelty. Like
which draws heaviliy from the existing creativity research to examine porpoises, it may produce something that is attributed with creativity,
the ostensible creativity of AI, leads to conclusion that the output of AI but actually it is at most mere novelty, and even this is not intentional.
qualifies as what has been called pseudo-creativity (Cropley, 1999; May, The same is true of another example of seemingly creative AI. Miller
1959; Nicholls, 1972). It is a particular kind of pseudo-creativity and (2019) described how, in the 1960s, an IBM supercomputer generated
may require that we follow the format which gave us “AI” (from artifi­ random lines instead of the graph that was expected, and the user “ran
cial intelligence) and refer to it as artificial creativity. down the hall shouting that the computer had produced art!” (p. 40)
Those lines may have been original, and even aesthetically pleasing, but
What is necessary for creativity but lacking in AI they were not intentionally drawn. Miller also described the famous
chess match between Garry Kasparov (who was World Champion) and
One answer to the question of whether or not AI can be creative is the IBM supercomputer, Deep Blue, in 1997. During the match “the
suggested by careful wording. This can add specificity to the way that computer came up with a sacrifice of such subtlety that Kasparov
the output of AI is described. More than once I have suggested that it is accused IBM of cheating.” It was not cheating. It also was not inten­
best to avoid the noun “creativity” because of its ambiguity (Runco, tionally subtle and strategy. It was glitch in the software.
2023b). Creative capacities are expressed in one way in childhood, but My point is that computers, and porpoises, may be viewed as creative
in a different way among mature and productive adults. They require when they are not. Their supposed creativity is not authentic creativity.
certain things in the arts but different things in other fields (e.g., science, It is a misattribution. It is very much like Eliza, a automated computer
entrepreneurship, IT, mathematics). Creativity is viewed in different psychotherapist (Weizenbaum, 1976) which comes across as authentic. I
ways in different cultures as well (Kharkhurin, 2012; Tan, 2016). The had several exchanges with Eliza as part of a class on AI when I was in
noun “creativity” does not apply equally in all cases and it thus makes college. It asked questions very much like a human non-directive hu­
sense to use the adjective “creative” instead. This compels specificity, as manistic psychologist. Eliza was especially convincing because
in “creative achievement,” “creative personality,” “creative behavior,” non-directive psychotherapists mostly ask questions, and the questions
“creative cognition,” and so on. are based on what the patient has expressed. Hence the label, “non-di­
Unfortunately this semantic suggestion does not apply to the term, rective” therapy. A computer can do much the same! The questions may
“creative AI.” That term would imply that AI can be creative, even be convincing–it can reword what has been expressed by a human
though it lacks some of the crucial characteristics of human creativity. patient–but the process is entirely computational and not at all mindful
Several important characteristics are discussed below. Before discussing and human. Yet it is easy to attribute mindfulness to Eliza during a
them it is useful to take a detailed look at what AI can do and the reasons conversation. Eliza was programmed before the chatbot named Eugene
that it is sometimes attributed with a creative capacity. Goostman fooled a human and passed the Turing Test for the first time.
To begin with, AI can be original, and its output is often useful. If AI (Simplifying some, the Turing Test asks a human to decide if he or she is
produces useful output, it may satisfy the requirement of “effectiveness” having a dialogue with a computer or a human.) The chatbot Goostman
(also called fit or “utility,” Tsao, Ting, & Johnson, 2019). But creativity fooled a human into believing that the chatbot was actually a teenage
is more than originality and effectiveness. Think for a moment about the Ukrainian boy.
research on divergent thinking. Tests of divergent thinking are usually The Lovelace Test, which hinges on a computer “originating an idea,”
scored for originality, but that does not make them tests of creativity. is more relevant to the discussion at hand. It focuses on the process
Originality is necessary but not sufficient for creativity. Originality can (originating) and not the product (output). Before discussing process in
be wildly ineffective, as is apparent in the divergent thinking of psy­ detail, it may help to take stock. The discussion so far has suggested that
chotic patients (Eysenck, 1997). Sometimes original things are original various things are lacking in AI. Intrinsic motivation, choice, and
precisely because they are worthless. Certainly divergent thinking tests intentionality are lacking, for instance, and each has been tied to crea­
provide useful estimates of the potential for creative thinking (Acar & tivity quite a few times (e.g., Gillebaart, Förster, Rotteveel, & Jehle,
Runco, 2020; Runco, 1991). They are estimates simply because, like all 2013; Horan, 2007; Pichot, Bonetto, Pavani, Arciszewski, Bonnardel, &
tests, they do not have perfect reliability. There is always some mea­ Weisberg, 2022; Runco, 1993, 1996, 2022). Intentionality implies
surement error. Most notable for the present purposes is that they do not choice and intrinsic motivation, and for parsimony, it was one of the two
capture all dimensions of creativity, even if they are scored for flexibility characteristics that was proposed in the updated standard definition of
and fluency, in addition to originality. The point here is that, just as creativity (Runco, 2023a). It characterizes human but not artificial
responses given to a divergent thinking question may be original, AI may creativity. At least as important is the authenticity of human creativity.
produce originality, but originality is far from synonymous with Authenticity is impossible for AI, but is also an important part of
creativity. human creativity (Kharkhurin, 2014; Maslow, 1973; Rogers, 1959). As
Several other examples of originality that is not creativity can be Maslow and Rogers described it, the authentic individual is
given. Consider the “creative porpoise” (Pryor et al., 1969). This was the self-accepting and, for that reason, does not excessively censor or
title of an article reporting research which used behavioral procedures constantly filter. The authentic individual expresses ideas and feelings
to “shape” novel behavior of porpoises. Shaping is the method of suc­ without manipulating them for the sake of others. It may be that AI can
cessive approximations. It begins with a behavior that is already in the avoid filters, but there is no self to express, so no possibility of
repertoire of an organism, and then gradually requires closer and closer authenticity.
approximations to a terminal behavior. In the case of the porpoises, the The definition of authenticity offered by the Stanford Encyclopedia of
terminal behavior was novelty. The porpoises in this research were Philosophy (2020a) is also very helpful. It defines “authentic” as "faithful
reinforced for emitting behaviors that they had not emitted before. to an original" and describes the authentic individual as “a person who
Those behaviors were in that sense novel. The cetaceans had no trouble acts in accordance with desires, motives, ideals or beliefs that are not
learning to emit novel behaviors. They are, after all, pretty darn smart. only hers (as opposed to someone else’s) but that also express who she
But they were not creative. What they emitted was mere novelty, lacking really is.” This captures some of the meaning of intrinsic motivation.
other required dimensions of authentic creativity. Then there is the Encyclopedia’s description of authenticity as diametric
The behavior of the porpoises was not intrinsically motivated. More to “derivative.” That would seem to apply to originality quite well. A
generally, the porpoises were not making intentional, mindful choices. derivative is quite unlikely to be original.
The novelty of the porpoises was elicited by trainers. Intrinsic motiva­ So far the discussion has identified criteria or dimensions of crea­
tion, mindfulness, and choice are enormously important parts of tivity that may distinguish human creativity from artificial creativity.

2
M.A. Runco Journal of Creativity 33 (2023) 100063

The more important question concerns “how” rather than “what.” A fully evaluate it. He put it this way:
consideration of how humans create–and in particular the process used
"A variant of Lady Lovelace’s objection states that a machine can
when creating–shows humans to use processes that are beyond AI. The
‘never do anything really new’. This may be parried for a moment
“how” is more important than identifying key characteristics because it
with the saw, ‘There is nothing new under the sun’. Who can be
offers an explanation of creativity. It is one thing to describe crea­
certain that ‘original work’ that he [sic] has done was not simply the
tivity–and the characteristics listed above help with that–but it is much
growth of the seed planted in him by teaching, or the effect of
more powerful to explain how creativity operates. Science needs ex­
following well-known general principles. A better variant of the
planations and cannot rely on mere descriptions. As Jay and Perkins
objection says that a machine can never ‘take us by surprise’....The
(1998) put it, there is a need to specify the mechanism that leads to
view that machines cannot give rise to surprises is due, I believe, to a
creative outcomes. Without specifying a mechanism, it is impossible to
fallacy to which philosophers and mathematicians are particularly
really understand creativity.
subject. This is the assumption that as soon as a fact is presented to a
mind all consequences of that fact spring into the mind simulta­
How do humans create?
neously with it. ...A natural consequence of doing so is that one then
assumes that there is no virtue in the mere working out of conse­
One view of the creative process (Runco & Chand, 1995) includes
quences from data and general principles.”
intrinsic motivation, and as noted above, that is not something that is
possible for a computer. That two-tiered process model, as well as the This is, for several reasons, an interesting point of view. First is
oft-cited model from Wallas (1926), also include problem finding, and Turing’s thinking about consequences. Consequences may be found
here again, this is not something we could reasonably expect of a through logic, and computers can explore logical results of premises and
computer. Problem finding is an umbrella term and includes problem situations. This is why King et al. (2007) were able to construct a robot
identification, problem definition, problem formulation, and problem scientist. It formulated hypotheses, but did so in a fashion that was
construction (Abdullah et al., 2020; Mumford et al., 1994; Runco, 1994). programmed and logical and not necessarily creative. Recall here the
A computer might find good problems but would have to be told to go criticisms of King et al.’s claims (Maffettone et al., 2023). Turing also
looking for them. This is the argument that Csikszentmihalyi (1988) brings surprise into the equation, and several scholars have suggested
used when he debated Simon (1988) about BACON, the problem solving that surprise is indicative of creativity (Bruner, 1979; Simonton, 2012).
software. Bruner equated effective surprise with creativity, and Simonton suggested
A recent version of this debate (King et al., 2009; Maffettone et al., that the non-obvious criterion used by the US Patent and Trademark
2023) focuses on the autonomous discovery by what King et al. called Office can be seen as a kind of surprise.
robot scientists. These robots ostensibly formulate and test hypotheses. Another interesting take on the Lovelace test was described by
Hypothesis formulation does sound like a kind of problem finding, but Marks (2022): “computer creativity will be demonstrated when a ma­
the support offered seems to focus on the autonomous conclusions rather chine’s performance is beyond the intent or explanation of its pro­
than the hypotheses. In any case, the robots were told to formulate grammer” (5th para). This fits nicely with the proposal above about
hypotheses. This dramatically detracts from their autonomy, at least intentionality. If it is “beyond the intent of the programmer,” it would
during the problem finding stage of the creative process. Actual problem have to be result of the intentions of the machine. “Beyond the expla­
finding on the part of AI would involve allowing them to scrape the web nation of the programmer” is also telling. If the programmer is unable to
for information, and then simply leaving them to their own devices (pun explain the ostensible creativity, the machine itself “ would seem to be
intended). If they find problems without being told to do so, nor told that originating the idea.”
problems should be identified, the debate might turn in the direction of Is it possible to have a creative product without using a creative
authentic problem finding. process? Perhaps by chance, but as suggested above, AI might use an
The product-process distinction used throughout the present dis­ uncreative process to produce something that is not really creative but is
cussion is a part of the widely-cited 4P and 6P frameworks (Rhodes, attributed with creativity. The same thing is true of intentionality: It
1961; Runco, 2007). The latter was formulated in part to distinguish makes no sense to believe that machines have intentions, but humans
between all of the performance approaches (e.g., Products, and another may attribute them with intentions. This possibility is entirely consistent
of the Ps, namely Persuasion) and all of the approaches focused on po­ with Kasof’s (1995) attributional theory of creativity with its depen­
tential (e.g., Personality, Place, and Process). Simonton (1988) proposed dence on social recognition. Versions of the attributional view are
that creative individuals and creative things change the way people sometimes accepted, as is evidenced by the fact that methods like the
think. They are in that sense persuasive. This was the fifth P and applies consensual assessment technical (CAT; Amabile, 1990) are popular. The
nicely to high-level, eminent creativity. The sixth was potential. Runco CAT and similar techniques rely on human judgment to decide what is
(1995) has previously suggested that persuasion, and views of creativity creative.
that require social recognition, conflate creative talent with other Critics of attributional theory argued that human judgment is far
non-creative things, such as impression management (Kasof, 1995), from perfectly reliable, and that it conflates the creativity of some
salience, contrarianism, and productivity. This is all relevant because outcome with the expectations and biases of whomever is doing the
much the same could be said about the output from AI: It only appears to judging (e.g., Policastro & Gardner, 1995; Runco, 1995). Not sur­
be creative. An attribution that AI is creative may not reflect creativity pringsly, groups of judges do not always agree with one another, which
and only creativity. raises the issue of generalizability, from one group of judges to other
Now we can go into more detail about the Lovelace test. It was groups (Runco, 1989; Runco, McCarthy, and Svensen, 1994). This is all
named after Ada Lovelace (1815-1852), daughter of Lord Byron, relevant because the labeling of AI creativity depends on judgments of
because she proposed that the critical test for a computer was whether or AI output. It is the output alone that is being attributed with creativity.
not it can "originate an idea." That of course puts the emphasis on Using Rhodes’ (1961) labels, the emphasis is on product and process is
originality, but it also implies that the underlying process is important. ignored.
After all, "to originate" describes the process used when something new
comes into being. This suggestion of process goes a long way towards Emergence, mirages, and hallucinations
answering the question of AI creativity. That is because AI can produce
things that are judged to be creative, but judgments are fallible, and, What of the process used by AI? It has access to enormous data bases
more importantly, AI does not use the same creative process as humans. and, for this reason, can find and then share reasonable answers to many
Turing (1950, pp. 450-451) reworded Lovelace’s position in order to questions (or prompts). In addition, AI can combine information from

3
M.A. Runco Journal of Creativity 33 (2023) 100063

different sources. This adds to the appearance of creative thinking, “My definition of the creative process is that it is the emergence in
which is often described as combinatorial (Bruner, 1979; Dietrich, 2015; action of a novel relational product, growing out of the uniqueness of
Poincare, 1905/1970). Bruner quoted Poincare on this point: creative the individual on the one hand, and the materials, events, people, or
thinking uses combinations that “reveal to us unsuspected kingships circumstances of his [her] life on the other....It makes no distinction
between… facts long known but wrongly believed to be strangers to one between ‘‘good’’and ‘‘bad’’ creativity. One man [sic] may be
another” (p. 5). Creative thinking may sometimes use some sort of discovering a way of relieving pain, whereas another is devising a
combinatorial process, but creative insights frequently go beyond new and more subtle form of torture for political prisoners. Both
existing information and the combination (and recombination) of it. these actions seem to me creative, even though their social value is
Hausman (1976) suggested that ideas and solutions that were in any way very different....Galileo and Copernicus made creative discoveries
dependent on already existing information, even metaphorically, are not which in their own day were evaluated as blasphemous and wicked,
truly creative. Interestingly, although he did not use the term “emer­ and in our day as basic and constructive" (quoted by Spooner, 2008,
gence”, he seemed to recognize its importance for the creative process p. 128)
when he wrote, "each appearance of genuine novelty is a sign of genuine
There is a great deal to unpack here, including the important
creativity....’Genuine novelty’ will refer to that character of the result of
distinction of malevolent and benevolent creativity (Cropley et al.,
a creative process which marks the result as different in kind or type from any
2010; McLaren, 1993; Runco, 1993) and the aforementioned blurring of
form available before the process before it began" (pp. 343-344). Hausman
discovery and creativity. What is most important for the present dis­
favored the descriptor spontaneity, which does make sense in that
cussion is Rogers’ reference to the “uniqueness of the individual.” This is
something arising spontaneously is not an obvious extension of what
not surprising, given Rogers’ view of creativity as inextricable from
came before.
self-actualization. When an individual is authentically self-expressive, it
The fact that AI relies on searches of existing data bases suggests that
is this uniqueness as an individual that is being expressed.
what it might be doing is discovering ideas and solutions. Urbina,
A similar assertion about the uniqueness of the individual as
Lentzos, Invernizzi, and Elkins (2022) were clear about this in their
expressed in creative work was recently implied by a novelist’s argu­
research on the drug discoveries by AI. Admittedly the distinction be­
ment that it is impossible for AI to be creative. Means (2023) wrote that
tween discovery and creation is a bit fuzzy (Boorstin, 1983; Root-­
Bernstein, 1988; Runco, 2023b). Martin and Wilson (2017) proposed “I’m not going to go pick a fight with A.I., or even argue with the fact
that all creativity is a kind of discovery. Still, there is an obvious logic to that this technology can mimic artwork, or assist humans in the
defining discovering as finding something (which is what AI does) and creation of art, but I can say...that A.I. will never be able to do what I
creation as bringing something new into being. can do because A.I. has never felt what I’ve felt. It will never move
How can ideas be found if they are not based on existing informa­ through the emotional matrix of living a singular, individual life.”
tion? There are several possibilities. One is nonlinear thinking (Richards,
A more graphic albeit ambiguous sentiment was expressed by a
2021; Schuldberg & Guisinger, 2021). Runco (2007) proposed some­
musician: “Eagles guitarist Joe Walsh, who once contributed to $28K
thing similar and described creative thinking as a kind of reasoning that
worth of damage at a Chicago hotel, doesn’t worry about AI because it
is not logical or rational in a traditional sense. He described creative
’can’t destroy a hotel room’” (Haroun, July 12, 2023).
cognition with an equation representing the process, with weights given
This position was foreseen by Jefferson (1949, quoted by Turing,
to each piece of relevant information. Creative thinking weighs origi­
1950, pp. 445-446): "Not until a machine can write a sonnet or compose
nality much more heavily than correctness, convention, and (tradi­
a concerto because of thoughts and emotions felt, and not by the chance fall
tional) logic. Another possibility is suggested by constructionistic theories
of symbols, could we agree that machine equals brain—that is, not only
(Piaget, 1976; Runco, 2003). Piaget, for example, described how the
write it but know that it had written it. No mechanism could feel (and
cognitive structures that allow understanding grow and progress to
not merely artificially signal, an easy contrivance) pleasure at its suc­
maturity. Perhaps the most intriguing process by which new ideas can be
cesses, grief when its valves fuse, be warmed by flattery, be made
found is emergence (Estes & Ward, 2002; Taylor, 1975; Richards, 2018;
miserable by its mistakes, be charmed by sex, be angry or depressed
Waller et al., 1993). Estes and Ward (2002) described how “novel fea­
when it cannot get what it wants" (italics added). This not only supports
tures are often attributed to a concept combination that are not attrib­
the thesis that humans use a unique creative process; it also is consistent
uted to either of its constituent concepts” (p. 149). Keenan, Stirrup, and
with the position that computers may produce things which are attrib­
Booth (2021).took a similar empirical approach and also reported the
uted with creativity, even when those outputs do not result from of a
emergence of novelty in conceptual thought. Richards (2018) described
creative process.
the role of emergence in human evolution and, of more direct relevance
There are claims that AI has demonstrated emergence. These claims
to the present argument, explained how emergence plays a substantial
have, however, been criticized. Schaeffer et al. (2023) demonstrated
role in human thought. She tied emergence to nonlinear thinking and
empirically that what appeared to be emergence by AI was actually a
creativity and was very clear that emergence is not just the appearance
mirage. A mirage is AI output that seems to be original but in reality is
of something new. Emergence indicates the appearance of something
not. Significantly, Schaeffer had data and compelling refutation of the
complex from simple premises or precedents. Richards stated, "each
supposed emergence. As a matter of fact, the claims about AI emer­
emergent phenomenon is, by definition, greater than the sum of its
gences may illustrate a thesis of the present article. In particular, the
parts—not explainable by a simple combination of what goes into it.
claims about emergences seem to be entirely descriptive. AI does
Hence, reductionist models (where all can be explained by the parts) will
something and programmers are not sure how it does it, so they label the
not work here" (2018, p. 2). The Stanford Encyclopedia of Philosophy
event an emergence. There is no explanation; this is merely descriptive.
(2020b) agrees about reductionism but introduces the interesting pos­
The data by Schaeffer et al., on the other hand, focus on mechanism.
sibility that there is a continuum, with some events and attributes
They tested AI emergences and found no support. The data do not
entirely dependent on existing components and pre-existing conditions
support emergence as an explanation for what are best viewed as mi­
and, at the other extreme, some things independent and entirely emer­
rages. The point here is that there is a difference between the explana­
gent. The quotation above from Hausman (1976) is applicable here in
tory power of data on mechanism vs the mere descriptions offered
that it referred to “a creative process which marks the result as different
without data on underlying processes. You might say that AI is attrib­
in kind or type from any form available before the process before it
uted with the capacity for emergences, but those attributions are not
began" (pp. 343-344).
based on data and are not explanations.
Rogers (1959, p. 71) described a slightly different kind of emergence:
Some erroneous AI output has been described as a kind of

4
M.A. Runco Journal of Creativity 33 (2023) 100063

hallucination. This is interesting because hallucinations are a part of can lead to unpredictable and even risky scenarios as these programs
some forms of (human) schizophrenia, and schizophrenia has in turn become more ubiquitous...[AI is] trained...by basically doing auto­
been associated with creativity (Michalica & Hunt, 2013). There is complete....We can’t look at how a person thinks and explain their
probably nothing to this, however, because the label “hallucination” reasoning by looking at the firings of the neurons. And it’s perhaps
applied to AI output is really just a loose analogy. The process that al­ even worse for these neural networks because we don’t even have the little
lows AI to hallucinate is nothing like the process humans experience bits of intuition that we’ve gotten from humans. We don’t really even
when they hallucination. That is, in a way, the general theme of this know what we’re looking for" (From Hassenfeld, 15 July 2023, italics
article. Humans use certain processes, computers use others. added).
The thesis of this article is tenable even with the simple premise that
Discussion
computer and human processes differ from one another. AI does not use
the same process as a human does when that human is thinking
One section of this article focused on characteristics and re­
creatively.
quirements of creativity. It suggested that AI can be original and effec­
Several times I stipulated that AI could be original and effective. This
tive, but it lacks other things that are a part of authentic creativity (e.g.,
should be qualified by the occurrence of hallucinations. They suggest
intrinsic motivation, intentionality, authenticity, problem finding). It
that even though AI is drawing from huge knowledge bases, it still makes
may be that intentionality covers intrinsic motivation, problem finding,
mistakes. Thus it is not always effective. Hiltzik (2023) recently listed
and choice. It may even cover authenticity, at least in the sense that an
some big mistakes, including AI providing lawyers with false precedents
authentic human regularly chooses (or intends) to express him- or her­
(which the lawyers actually used–and for which they were heavily
self (rather than conform).
fined); a Texas professor flunking his entire class based on AI’s falsely
The other major section of this article focused on the distinction
identifying everything as plagiarism; and researchers at MIT reporting
between process and product. It brought another one of the Ps into the
(erroneously) that AI could earn a perfect score on the math and com­
discussion as well. This was persuasion, which is a part of the theory that
puter science examinations.
creativity must be defined in terms of social recognition. That view
In some ways the deficiencies of AI–including those leading to hal­
applies to AI, or may, if humans misjudge such that they attribute
lucinations and mirages, as well as the impossibility of authenticity,
creativity to the output from AI, just because it is original and effective.
intentionality, spontaneity, intrinsic motivation, and problem fin­
Such output may also surprise a human audience (“how could a machine
ding–do not matter. As mentioned above, businesses will promote AI as
do that?”). Surprise is one part of some definitions of creativity but is not
creative even if it lacks authenticity and the other requirements of
sufficient in and of itself (Simonton, 2012).
creativity. Huge investments are being made into AI (e.g., Microsoft’s
The two sections of this article, one focusing on characteristics and
$10 billion given to OpenAI), and the companies doing the investing
the other focusing on processes, are complementary. The characteristics
would not risk such a huge amount without confidence that there will be
influence the process and result from it. Authenticity, for example, leads
a return. Very likely, the companies which developed AI are primarily
the individual to a particular kind of honest and relatively uninhibited
concerned about the output (i.e., the product). They probably do not
self-expression; and self-expression is a process. Similarly, intentionality
care if artificial creativity uses a process that differs dramatically from
is apparent in the individual’s motives and choices, and these direct the
that used by humans when they are creative. Scientists studying crea­
creative process.
tivity probably care because their objective is to understand creativity,
The emphasis in this article was on process and product in part
and as mentioned above, this means that the process is key. Businesses,
because they differ in their explanatory power. To really explain crea­
on the other hand, care about the bottom line, and that can be enriched
tivity it is necessary to understand the mechanism that produces it (Jay
with output that gives the impression of creativity even if it is not truly
& Perkins, 1997). With this in mind it is vital to distinguish between
creative.
human and artificial creativity. Only then can we understand the
Add to this the fact that AI can change the world without authentic
mechanism that underlies and allows creativity. Only then can will
creativity. This is frightening: An uncreative AI can do great damage.
recommendations for fulfilling potential have a chance of succeeding.
Think about all of the thousands of computer viruses that are developed
An updated standard definition would be helpful in this regard (Runco,
each year. What if the developers of malware turn their skills to AI?
2023a), but that is only one step towards recognizing creativity that is
Fritsch, Jaber and Yazidi (2022) detailed how this has already
authentic and that which is artificial.
happened! They wrote, “artificial intelligence (AI) and machine learning
The focus on process does not imply that the processes used by
(ML) methods are increasingly adopted in cyberattacks. AI supports the
humans are well understood. I am well aware that the human mind was
establishment of covert channels, as well as the obfuscation of malware.
called a Black Box well before that label was used to describe AI.
Additionally, AI results in new forms of phishing attacks and enables
Nonetheless, the human process may very well be better understood
hard-to-detect cyber-physical sabotage. Malware creators increasingly
than the process used when AI produces original output. Consider all of
deploy AI and ML methods to improve their attack’s capabilities.”
the process research on creativity! Also consider the admission of Sam
There are implications for programs intended to support or even
Bowman, AI scientist at NYU: "If we open up ChatGPT or a system like it
enhance creativity, in addition to the implications for the standard
and look inside, you just see millions of numbers flipping around a few
definition used in research and theory. Creativity programs always as­
hundred times a second...and we just have no idea what any of it means"
sume a definition of creativity, and they should probably recognize the
([Link]
things that are required for authentic creativity rather than merely
gpt-ai-science-mystery-unexplainable-podcast). This is part of a series
satisfying the criteria of artificial creativity and the existing (and inad­
titled, "AI: The Black Box." The degree of uncertainty about what is going
equate) standard definition.
on in AI may come as a surprise, given that it was build by humans. Here
The stipulation mentioned above that AI may be original and effec­
is how an interview with Bowman summarized the situation:
tive should also be qualified by the fact that its originality is derivative
"ChatGPT runs on something called an artificial neural network, and dubious, and it lacks several of the characteristics of authentic
which is a type of AI modeled on the human brain. Instead of having (human) creativity. AI derives its output by “scraping text on the web”
a bunch of rules explicitly coded in like a traditional computer pro­ (Hiltzik, 2023). Thus it can only produce artificial creativity. Early in
gram, this kind of AI learns to detect and predict patterns over this article the term “creative AI” was rejected. The term artificial crea­
time.... [But] because systems like this essentially teach themselves, tivity makes more sense, at least because the concept of artificial intel­
it’s difficult to explain precisely how they work or what they’ll do. Which ligence is so widely used, and “artificial creativity” is aligned with

5
M.A. Runco Journal of Creativity 33 (2023) 100063

“artificial intelligence.” Given that artificial creativity lacks much of Maffettone, P. M., et al. (2023). What is missing from autonomous discovery? Open
challenges for the community. Arxiv arXiv, 2304, 11120. [Link]
what is expressed in human creativity, and it uses wildly different pro­
arXiv.2304.1112
cesses, it is most accurate to view the ostensibly creative output of AI as a Marks, R. J. (2022). Did the GPT3 chatbot pass the Lovelace Test? Mind Matters. https:
particular kind of pseudo-creativity. //[Link]/2022/12/did-the-gpt3-chatbot-pass-the-lovelace-test/.
Martin, L., & Wilson, N. (2017). Defining creativity with discovery. Creativity Research
Journal, 29, 417–425. [Link]
Declaration of Competing Interest Maslow, A. H. (1973). Creativity in self-actualizing people. In A. Rothenberg, &
C. R. Hausman (Eds.), The creative question (pp. 86–92). Durham, NC: Duke
No conflicts. University Press.
May, R. (1959). The nature of creativity. Review of General Semantics, 16, 261–276.
[Link]
References McLaren, R. B. (1993). The dark side of creativity. Creativity Research Journal, 6,
137–144.
Abdulla, A. M., Paek, S. H., Cramond, B., & Runco, M. A. (2020). Problem-finding and Means, D. (2023). A.I. Can’t write my cat story because it hasn’t felt what I feel. New
creativity: A meta-analytic review. Psychology of Aesthetics, Creativity, and the Arts, York Times. [Link]
14, 3–14. [Link] Michalica, K., & Hunt, H. (2013). Creativity, schizotypicality, and mystical experience:
Acar, S., & Runco, M. A. (2019). Divergent thinking: New methods, recent research, and An empirical study. Creativity Research Journal, 25, 266–279. [Link]
extended theory. Psychology of Aesthetics, Creativity, and the Arts, 13, 153–158. 10.1080/10400419.2013.813780
[Link] Miller, A. I. (2019). The Artist in the Machine: The World of AI-Powered Creativity.
Albert, R. S. (1996). What the study of eminence can teach us. Creativity Research Journal, Cambridge, MA: MIT Press.
9, 307–315. Mumford, M. D., Reiter-Palmon, R., & Redmond, M. R. (1994). Problem construction and
Amabile, T. M (1990). Within you, without you: The social psychology of creativity, and cognition: Applying problem representations in ill-defined domains. In M. A. Runco
beyond. In M. A. Runco, & R. S. Albert (Eds.), Theories of creativity (pp. 61–91). (Ed.), Problem finding, problem solving, and creativity (pp. 3–39). Norwood, NJ: Ablex.
Newbury Park, CA: Sage. Nicholls, J. G. (1972). Creativity in the person who will never produce anything original
Boorstin, D. J. (1983). The discoverers: Man’s search of his world and himself. New York: and useful. The concept of creativity as a normally distributed trait. American
Random House. Psychologist, 27, 717–727.
Bruner, J. (1979). The conditions of creativity. In J. Bruner (Ed.), On knowing: Essays for Piaget, J. (1976). To understand is to invent. New York: Penguin.
the left hand (pp. 1–30). Boston: Harvard University Press. Pichot, N., Bonetto, E., Pavani, J. B., Arciszewski, T., Bonnardel, N., & Weisberg, R. W.
Cropley, A. J. (1999). Definition of creativity. In M. Runco, & S. Pritzker (Eds.), (2022). The construct validity of creativity: Empirical arguments in favor of novelty
Encyclopedia of creativity (pp. 511–524). San Diego, CA: Academic Press. as the basis for creativity. Creativity Research Journal, 34, 2–13. [Link]
Cropley, D. H., Cropley, A. J., Kaufman, J. C., & Runco, M. A. (Eds.). (2010). The dark side 10.1080/10400419.2021.1997176
of creativity. New York: Cambridge University Press. Plain, C. (2023). Artists beware: “Game changer” test results show AI is more creative
Csikszentmihalyi, M. (1988). Solving a problem is not finding a new one: A reply to than 95% of humans. The Debrief. [Link]
Herbert Simon. Journal New Ideas in Psychology, 6, 183–186. anger-test-results-show-ai-is-more-creative-than-99-of-humans/.
Dietrich, A. (2015). How creativity happens in the brain. New York: Palgrave MacMillan. Poincare, H. (1970). Mathematical creation. In P. E. Vernon (Ed.), Creativity. New York:
Edwards, B. (2023). OpenAI execs warn of “risk of extinction” from artificial intelligence Penguin.
in new open letter. Ars Technica. [Link] Policastro, E., & Gardner, H. (1995). Naive judgment and expert assessment: A critique of
2023/05/openai-execs-warn-of-risk-of-extinction-from-artificial-intelligence-in-ne the attributional perspective. Creativity Research Journal, 8, 391–395. [Link]
w-open-letter/. org/10.1207/s15326934crj0804_5
Estes, Z., & Ward, T. B. (2002). The Emergence of Novel Attributes in Concept Pryor, K. W., Haag, R., & O’Reilly, J (1969). The creative porpoise: Training for novel
Modification. Creativity Research Journal, 14, 149–156. [Link] behavior. Journal of the Experimental Analysis of Behavior, 12, 653–661.
S15326934CRJ1402_2 Rhodes, M. (1961). An analysis of creativity. Phi Delta Kappan, 42, 305–310.
Eysenck, H. J. (1997). Creativity and personality. In M. A. Runco (Ed.), The creativity Richards, R. (2018). Emergence of life and creativity. In R. Richards (Ed.), Everyday
research handbook (pp. 41–66). Cresskill, NJ: Hampton. Vol. 1. Creativity and the Healthy Mind: Dynamic New Paths for Self and Society. London:
Fritsch, L., Jaber, A., & Yazidi, A (2022). An overview of artificial intelligence used in Palgrave MacMillan. [Link] [Link]
malware. In E. Zouganeli, A. Yazidi, G. Mello, & P. Lind (Eds.), Nordic Artificial [Link]/chapter/.
Intelligence Research and Development. NAIS 2022. Communications in Computer and Richards, R. (2021). Three Cs revisited–Chaos, complexity, and creativity: where
Information Science. Springer. [Link] vol. nonlinear dynamics offers new perspectives on everyday creativity. Journal of
1650. Creativity. [Link]
Gillebaart, M., Förster, J., Rotteveel, M., & Jehle, A. C. M. (2013). Unraveling effects of Rogers, C. R. (1959). Toward a theory of creativity. In H. H. Anderson (Ed.), Creativity
novelty on creativity. Creativity Research Journal, 25, 280–285. [Link] and its cultivation (pp. 69–82). New York: Harper & Row.
10.1080/10400419.2013.813781 Root-Bernstein, R. (1988). Discovering. Cambridge University Press.
Haroun, A. (2023). Eagles guitarist Joe Walsh, who once contributed to $28K worth of Runco, M. A. (1989). The creativity of children’s art. Child Study Journal, 19, 177–189.
damage at a Chicago hotel, doesn’t worry about AI because it ’can’t destroy a hotel Runco, M. A. (1991). Divergent thinking. Norwood, NJ: Ablex Publishing Corporation.
room. Insider. [Link] Runco, M. A. (1993). Moral creativity: Intentional and unconventional. Creativity
stroy-a-hotel-room-2023-7#:~:text=Joe%20Walsh%2C%20the%20guitarist%20for, Research Journal, 6, 17–28.
%2C%22%20Walsh%20told%20the%20AP. Runco, M. A. (1994). Conclusions concerning problem finding, problem solving, and
Hassenfeld, N. (15 July 2023). Even the scientists who build AI can’t tell you how it creativity. In M. A. Runco (Ed.), Problem finding, problem solving, and creativity (pp.
works. "We built it, we trained it, but we don’t know what it’s doing." [Link] 272–290). Norwood, NJ: Ablex.
[Link]/unexplainable/2023/7/15/23793840/chat-gpt-ai-science-mystery-une Runco, M. A. (1995). Insight for creativity, expression for impact. Creativity Research
xplainable-podcast. Journal, 8, 377–390.
Hausman, C. R (1976). Creativity and rationality. In A. Rothenberg, & C. R. Hausman Runco, M. A. (1996). Personal creativity: Definition and developmental issues. New
(Eds.), The creativity question (pp. 343–351). Durham, NC: Duke University Press. Directions for Child Development, (72), 3–30. Summer.
Hiltzik, M. (2023, July 16). Spreading faster than AI chatbots: The hype. Los Angeles Runco, M. A. (2003). Education for creative potential. Scandinavian Journal of Education,
Times. [Link] 47, 317–324.
-eec1-4447-8ba0-450881a594c2. Runco, MA (2007). A hierarchical framework for the study of creativity. New Horizons in
Horan, R. (2007). The relationship between creativity and intelligence: A combined Education, 55(3), 1–9.
yogic-scientific approach. Creativity Research Journal, 19, 179–202. [Link] Runco, M. A. (2022). Positive creativity and the intentions, discretion, problem finding,
10.1080/10400410701397230 and divergent thinking that support it can be encouraged in the classroom.
Jay, E., & Perkins, D. (1997). Creativity’s compass: A review of problem finding. In M. A. Educational Science, 12(5), 340. [Link]
Runco (Ed.), Creativity research handbook (vol. 1, pp. 257-293). Cresskill, NJ: Runco, M. A. (2023a). Updating the standard definition of creativity to account for the
Hampton Press. artificial creativity of AI. Manuscript under review.
Jefferson, G. (1949). The mind of mechanical man. British Medical Journal, 1105–1110. Runco, M. A. (2023b). Creativity: Research, development, and practice (3rd ed.). San Diego,
Kasof, J. (1995). Explaining creativity: The attributional perspective. Creativity Research CA: Academic Press.
Journal, 8, 311–366. Runco, M. A. (2023c, May). Presentation at the Annual Southern Oregon University
Keenan, M., Stirrup, L., & Booth, N. (2021). The emergence of combinations of behavior Creativity Conference, Ashland, Oregon, USA.
in an equivalence class without explicit training of a function. Mexican Journal of Runco, M. A. (2023d). AI and Creativity. April. Pittsburgh PA: Keynote Address to BEST
Behavior Analysis, 47. [Link] Robotics.
Kharkhurin, A. V. (2014). Creativity 4in1: Four-Criterion Construct of Creativity. Runco, M. A., & Chand, I. (1995). Cognition and creativity. Educational Psychology
Creativity Research Journal, 26, 338–352. Review, 7, 243–267.
King, R. D., Rowland, J., Oliver, S. G., Young, M., Aubrey, W., Bryne, E., Liakata, M., Runco, M. A., Johnson, D., & Gaynor, J. R. (1999). The judgmental bases of creativity
Markham, M., Pir, P., & Clare, A. (3 April 2009). The automation of science. Science, and implications for the study of gifted youth. In A. Fishkin, B. Cramond, &
324, 85–89. [Link] P. Olszewski-Kubilius (Eds.), Investigating creativity in youth: Research and methods
(pp. 113–141). Cresskill, NJ: Hampton Press.

6
M.A. Runco Journal of Creativity 33 (2023) 100063

Runco, M. A., McCarthy, K. A., & Svensen, E. (1994). Judgments of the creativity of Tan, C. (2016). Creativity and confucius. Journal of Genius and Eminence, 1, 79–84.
artwork from students and professional artists. Journal of Psychology, 128, 23–31. [Link]
Schaeffer, R., Miranda, B., & Koyejo, S. (2023). Are emergent abilities of large language Taylor, I. A., & Taylor, I. A. (1975). An retrospective view of creativity investigation. In
models a mirage? arXiv, 2304, 15004v2. [Link] J. W. Getzels (Ed.), Perceptives in creativity (pp. 1–36). Chicago, IL: Aldine.
arXiv.2304.15004 [[Link]]. Tsao, J. Y., Ting, C. L., & Johnson, C. M. (2019). Creative Outcome as Implausible Utility.
Schuldberg, D., & Guisinger, S. (2021). Creativity and nonlinear dynamical systems. Review of General Psychology, 23, 279–292.
Journal of Creativity. [Link] Turing, A. M. (1950). Computing machinery and intelligence. Mind, LIX, 433-360.
Spooner, M. T. (2008). Commentary on malevolent creativity. Creativity Research Journal, Urbina, F., Lentzos, F., Invernizzi, C., & Elkins, S. (2022). Dual use of artificial-
20, 128–129. [Link] intelligence-powered drug discovery. Nature Machine Intelligence, 4, 189–191.
Stanford University Encyclopedia of Philosophy. (2020a). Authenticity. [Link] [Link]
[Link]/entries/authenticity/ First published Thu Sep 11, 2014; substantive revision Wallas, G. (1926). The art of thought. New York: Harcourt Brace & World.
Thu Feb 20, 2020 Accessed 5 June 2023. Waller, N. G., Bouchard, T. J., Lykkens, D. T., Tellegen, A., & Blacker, D. M. (1993).
Stanford University Encyclopedia of Philosophy. (2020b). Emergent properties. [Link] Creativity, heritability, familiality: Which word does not belong? Psychological
[Link]/entries/properties-emergent/. Inquiry, 4, 235–237.
Sternberg, R . J. (2000). Creativity is a decision. In A. L. Costa (Ed.), Teaching for Weizenbaum, J. (1976). Computer power and human reason: From judgment to calculation.
intelligence II (pp. 85–106). Arlington Heights, IL: Skylight Training and Publishing. San Francisco, CA: Freeman.