Abstract

"The most frequently expressed of the new distinctions uses emotions

to draw a line between computers and people" (Turkle, 2005, p. 63).

In October 2018, the auction house Christies sold the painting ‘Portrait of Edmond Belamy’ for $432,500. This is a common price for art, and both the artwork—a portrait of a man in a loose expressionist style reminiscent of Francis Bacon (Fig. 1)—and the purchase would typically not be all that remarkable. Except that, this portrait was created by an artificial intelligence (AI) algorithm by the art group "Obvious" (https://obvious-art.com). Reactions were diverse, but often negative. As put by Jerry Saltz (2018, para.1), epitomizing especially the art-critical distaste about the price and the selling of the picture, "an artwork made by Artificial Intelligence just sold … I am shocked, confused, appalled".

Explaining this shock and the disappreciation of the artwork, two common themes emerged as to what was missing in such pieces: the intentions of the artist and the ensuing emotional engagement on both the producer and the perceiver's side. A core aspect of making art seems to be the ability to record and then impart some aspects of a creator's reality, their views, and their feelings to an audience. Especially this latter aspect, often in the form of specific emotions, is suggested to be a main aspect of such connections and a core element—as a human communicative act—of art's definition itself. Tolstoy put it this way: "to evoke in oneself a feeling one has once experienced, and… then, by means of movements, lines, colors, sounds, or forms … to transmit that feeling that others may experience the same feeling—this is the activity of art" (1898/1962, p. 68, italics added).¹ As viewers, we are tasked with completing these connections—anticipating, perceiving, feeling into, and decoding the mind and the feelings of the artist (Collingwood, 1938; Donald, 2006; D'Inverno & Mccormack, 2015; Gerger et al., 2018; Hertzmann, 2018; Langer, 1954). This basic act of recognizing that others have emotional experiences and expecting that we as viewers might be able to, or even that we should, share these with them, can be connected to discussions of empathy, or the capacity to feel and understand what another human is thinking or experiencing (Hardee, 2003; Singer & Lamm, 2009; see also Gerger, Pelowski & Leder, 2017). Such an empathic-emotional expectation and processing basis is even argued to be a necessary precursor, in human evolution, of our use of images, media, or of artmaking itself (Donald, 2006; Dutton, 2000; Gell, 1998; Noë, 2015). Even in competing theories downplaying the art as expressionists' idea,² there is still a core acceptance that feelings could, or may often, be communicated due to this inescapable core human covenant.

Because it seems problematic to ascribe such intentions to an AI or to any computer, they, thus, cannot be presumed to actively transmit emotions to a viewer. Nor are there really any specific ‘true’ emotions to be evoked or received. Therefore, art made by AI can, in the critics' view, not be art (Jones, 2018; Kreye, 2018). Rather, without an artist to even begin the communicative act, the entire interaction becomes an empty gesture—a false covenant, an "illusion of art" (Hoel, 2022, par.4)—without the rich emotional connection that makes art the vibrant human medium that it, presumably, often is. This argument, of course, is not only confined to the sale of six-figure Fine Art examples but is a common, and currently increasing theme, when describing how humans engage with many forms of contemporary media (e.g., Lawson-Tancred, 2022; Plunkett, 2022a, 2022b),³ and also argued to be a core defining difference in our engagements with computers (Turkle, 2005). However, is it really true that humans do not look for and feel emotions even from artworks that they know are computer generated? A good deal of theory, and also some empirical evidence, suggests that looking for, finding, and even feeling, emotions and ascribing intentionality may be a general human trait (e.g., Currie, 2011; Gombrich, 1963; Hong, 2018; Lu, 2005). Humans may have a tendency to see agency in many objects whether or not—or rather, as if—they were from, or of, humans (e.g. Aggarwal & McGill, 2007; Guthrie, 1995; Müller et al., 2018; Waytz et al., 2010). Empathic and emotional engagements, may also be particularly strong with computers, and especially AI, which enable interaction characterized by responsiveness and a seemingly specific reference to the individual (Lee et al., 2006; Turkle, 2005; see also Luscombe, 2022). This sets up a quite simple set of questions that have nonetheless not been empirically assessed. Critical arguments aside, do we find emotional connections in the products of AI ‘artists,’ even if individuals are told that such art is by a computer? Might we even find intentionality tied to these responses? Answers to these questions have a practical importance as AI-derived art or other forms of expression—in movies, animations, social media images (e.g., Datta & Goswami, 2021; Plunkett, 2022a, 2022b)—have taken on an increasing relevance with major economic, cultural, and social implications (Bickley et al., 2022, pp. 2055–2084; Giordano et al., 2021; Ransbotham et al., 2021) and become more and more an aspect of our everyday media existence. Answers also have theoretical importance as they may reveal something fundamental about the nature of empathy, intentionality, and especially emotional connections in our engagement with the environment as well as AI itself.

We can also ask, what of the actual artistic product? As art by humans continuously comes to resemble the products of computers, and computer products look more and more human, can we identify when something is from a human, even when artworks might look similar to computer generated examples? To what extent is this dependent on our beliefs regarding the provenance (AI or human) of the artwork? In which ways does context contribute to our ascriptions of intentions and influence our emotional responses (Hong, 2018; Köbis & Mossink, 2021)? Setting aside the very real economic and legal (e.g., copyright, see Rose, 2022; Vincent, 2022) implications of AI-derived products, it is vital to continue to understand how we are in fact engaging the multiple media around us, and what sorts of psychological reactions or processes might be routine, or different (Fingerhut, 2021). Despite this interest, the actual investigation of our engagements with AI art, specifically involving intentionality, differing provenance, and nuanced emotions, is only now emerging.

To explore this, we conducted an empirical study of emotion and intention perception in viewers of art made either by humans or a computer. The study considered the aspects of felt and intuited emotions, intentions, and also the actual aspects of emotion communication between viewers and human artists when art is both framed as computer or human generated. We begin with a review of the literature on emotion sharing and communication in art as well as an assessment of the main aspects of human-computer interactions involving artworks. We then will survey studies on the general topic of labeling, provenance, and past studies investigating interactions with computer media that were the basis for our study, which serves as one of the first explorations of nuanced intended and felt emotional experience in human-computer art interactions.

References (88)

Cited by (0)

Recommended articles (0)