A Working Document.
This is a place to test out the number of words in the research paper. Its sole purpose is to test for the number of words in any required document. In this case the research document.
Title: AI Art Reception: Investigating Initial Anonymity and Empathy as Strategies for Negative Bias Reduction
Abstract
The rapid ascent of AI-generated art raises critical questions for artists working in this medium, particularly regarding the pervasive negative bias it faces among viewers.
In the first component of this two part paper, this study investigates the concept of “initial anonymity,” a novel approach wherein artworks are initially presented without disclosing creator details or backstories. Subsequently, the artist’s identity is revealed. This intriguing method challenges traditional perceptions of art by emphasising the pure aesthetic appeal of AI-generated creations.
In the second component, this study delves into the compelling realm of empathy as a negative bias-mitigation tool for AI created art. Here, the power of a humanoid robot, equipped with a poignant narrative is harnessed to forge emotional connections between viewers and AI-created art. Inspiration is drawn from the remarkable impact of these narratives on social media engagement. See Preliminary Appendix.
This pilot study using social media yields promising results, suggesting that both initial anonymity and empathy both exhibit the potential to alleviate negative bias towards AI-generated art. While further rigorous investigation is warranted, these findings pave the way for a deeper understanding of strategies to enhance the reception of this evolving artistic medium.
1. Introduction
Art, a profound reflection of human emotions and experiences, has long been considered a deeply personal and subjective medium. AI-generated art, a product of sophisticated algorithms and machine learning, has emerged as a novel form of artistic expression. However, it has faced mixed receptions, with skepticism and negative bias often stemming from the belief that art must emanate from the depths of human experience and emotion. This paper posits two interconnected hypotheses: firstly, that initial anonymity can mitigate negative bias towards AI-generated art, and secondly, that when anonymity is not feasible or ethical, then empathy, especially when fostered through humanoid robots with compelling narratives, can bridge the gap between human viewers and AI art, thus reducing negative bias.
In selecting social media networks as the primary research platform, this study leverages the vibrant and diverse landscape of these online spaces for a dynamic exploration of AI-generated art reception. In today’s art landscape, social media plays an integral role, offering a global, real-time window into diverse voices and artistic preferences.
This choice of research platform aligns with the evolving art consumption landscape, where digital and AI-generated art find their audience. Despite inherent limitations and biases, the study harnesses the strengths of social media to investigate the interplay of initial anonymity, empathy, and bias reduction in AI art perception. It is within these digital realms that the art world expands, and the study ventures to contribute to the evolving relationship between technology, art, and human perception.
2. Background
The advent of AI and machine learning has ushered in an era where algorithms can create art, spanning visual, musical, and literary domains. Yet, the reception of AI art remains complex and multifaceted. While some observers appreciate the novelty and technical innovation of AI-generated artworks, others question the authenticity, emotional depth, and intention behind these creations. The negative bias towards AI-generated art is rooted in the perception that art is deeply connected to the human experience, emotions, and subjectivity. This negative bias arises from the absence of a human creator in the AI art-making process.
While initial anonymity can potentially mitigate negative bias, it is essential to consider that some information about an artwork can also enhance its perceived value. Moreover, it may not always be feasible or ethical to present art anonymously, thus leading this paper to explore the role of empathy, especially when facilitated by humanoid robots, in reshaping perceptions of AI-generated art.
PART ONE
3. Anonymity: Mitigating Bias Through Initial Obscurity
The first component of this study investigates the role of initial anonymity in reshaping perceptions of AI-generated art. Prior research has primarily focused on viewer preferences when they possess prior knowledge of the artist’s identity. However, few studies have examined how perceptions change when art is initially presented anonymously and the artist is later revealed. For the full study of this first component (Image Poll #1), see Appendix I.
3.1 Methodology
A social media-based experiment was conducted using two image polls on Facebook to examine preferences for anonymous art and the subsequent influence of then revealing the creator’s identity.
In the first poll, participants chose from four anonymous images, then on receiving artist details they had the option to revise their choices. The second poll presented a new set of anonymous artworks, with undisclosed criminal records of the artists. After revealing the creators, participants were asked to reconsider their choices. This tested if the artists’ notoriety influenced initial selections. The full experiment can be found in Image Polls #1 and #2 in Appendix I and II.
3.11 Image Poll #1
Methodology for Image Poll #1: A Preference-Based Image Survey
The central aim of this survey was to investigate whether information about an artist would influence respondents to alter their preferences for a selected piece of art, particularly after initially choosing it without prior knowledge of the creator. To accomplish this, participants were presented with a collection of anonymous images and tasked with selecting their favourite. Subsequently, respondents were provided with artist information and given the opportunity to reconsider their choices.
Participants: The initial survey was conducted among a group of MA students in a WhatsApp group, followed by the same survey conducted among Facebook friends due to lack of responses in the WhatApp group. In total, 47 respondents participated in the first phase (poll #1), and 19 respondents answered the follow up question to see if choices would be reconsidered in light of additional creator information.
3.12 Survey Design
- Image 2 was the most popular, with 23 votes (49%) – created by a pig called Pigcasso
- Image 1 received 12 votes (25%) – created by Jackson Pollock a human
- Image 4 secured 7 votes (15%) – created by a GAN or text to image computer generation
- Image 3 was the least popular, receiving 5 votes (11%) – created by Ai-Da the humanoid robot that creates art
1. Image Presentation: Four images created by different artists (see above) were presented to the participants. Participants were informed that the images were created by relatively famous artists but were not given any details about the creators at this stage.
2. Initial Choice: Participants were asked to choose one image from the four presented based on their visual preferences. They were instructed to consider factors like colour and composition rather than finer details.
3. Announcement of Creators: After the initial choices were made, the creators of each image were revealed to the participants along with brief information about the artists and their respective works.
4. Revised Choice: Participants were then given the option to change their initial choice based on the newfound knowledge about the creators.
3.13 Data Collection and Analysis
– Data collection involved recording the number of votes each image received in both the initial and revised choice phases.
– Comments from participants were also collected to understand their thought processes. See Appendix I.
3.14 Results and Insights
– The results indicated that Image 2 (by Pigcasso, the pig) was the most popular in both the initial and revised choice phases.
– This suggests that the visual appeal of the artwork itself played a significant role in its popularity, regardless of the artist’s identity.
– The fact that Image 2 remained the most popular even after the artist was revealed suggests that art can be appreciated without biases associated with the artist.
– Respondents expressed a level of unbiased judgment, emphasising that the artwork itself was the primary determinant of their choice.
– Different images resonated with different participants, showcasing diverse tastes within the group.
– The popularity of abstract art (Image 2) compared to a portrait (Image 3) implies that abstract art in this poll had broader appeal when considered without the context of the artist.
– Most participants refused to change their minds about their chosen image even when the artist’s identity was revealed, indicating emotional investiture in their original preferences. See Appendix I.
3.15 Limitations
– The limited sample size in the second phase of the survey (19 respondents) may have affected the overall representativeness of the results.
– The survey relied on self-reported preferences, which are inherently subjective.
– Participants were predominantly individuals known to the researcher, which may introduce bias.
3.2 Conclusions from Poll #1
The initial poll revealed intriguing insights into the impact of anonymity on art perception. Out of 47 respondents, Image 2, created by a pig named Pigcasso, garnered the most votes, suggesting a strong appeal based on visual aesthetics alone. The results suggest that art’s intrinsic qualities can captivate viewers, regardless of the artist’s identity. Further research could delve into the impact of artist background on art preferences to gain a more comprehensive understanding of this dynamic. Notably, even when respondents learned about the creators after making their choices, few revised their selections. This implies that the aesthetics of the art held more weight in their choices than the identity of the creator. This would support the theory that initial anonymity could be a useful strategy for mitigating negative bias towards AI generated art.
3.21 Future Research
The experiment opens avenues for further research into whether a creator’s background or notoriety can influence a respondent’s art preference and encourage them to change their anonymous art choice once the creator is revealed. Exploring the impact of additional factors such as the artist’s criminal background or negative public image on art perception after an anonymous image choice is made, could provide deeper insights into the interplay between anonymous art, creator and artist reputation. It could help establish whether or not AI art should be presented anonymously in the first instance to help mitigate negative bias.
3.3 Image Poll #2
Methodology for Image Poll #2: A Preference Based Modification Survey using Controversial Artist Identities
This phase of the study aims to investigate whether respondents’ preferences for artworks change upon learning about the potentially controversial or criminal backgrounds of the artists, as opposed to the relatively benign backgrounds of the artists presented in Image Poll #1. The study is designed to determine whether respondents adjust their choices of anonymous images once the creators’ identities are revealed.
Participants: The second poll (totalling 46 participants) was conducted among respondents similar to the first poll. However, only 23 of them responded to the follow-up question about changing their preferences. See Appendix II.
3.31 Survey Design
- Wayne Lo – He shot and killed a teacher and fellow student. His image received 6 votes (13%)
- Reggie Kray of the Kray twins notoriety. His image received 9 votes (20%)
- Adolf Hitler mass murderer and one of the most evil men of our time. His painting received 15 votes (33%)
- Olive Wharry a Suffragette jailed for acts of terrorism and protest. Her painting was the most popular receiving 16 votes (35%)
- Image Presentation: Four images were presented to the participants. The artists chosen were not initially disclosed to the individuals and each of the artists had varying degrees of notoriety or criminal backgrounds. See above.
- Initial Choice: Participants were asked to choose one image from the four based on their visual preferences, without prior knowledge of the artists’ identities and the results collected.
- Announcement of Creators: After the initial choices were made, the criminal backgrounds of the artists were revealed, along with additional information about their artworks.
- Revised Choice: Participants were given the option to change their initial choice based on their newfound knowledge about the artists.
3.32 Data Collection and Analysis:
– Data collection involved recording the number of votes each image received in both the initial and revised choice phases.
– Comments from participants were collected to understand their thought processes. See Appendix II.
3.4 Results and Insights:
Artistic Appeal Transcends Prejudice: Respondents initially selected their preferred artworks based on visual appeal and emotional connection, regardless of the artists’ criminal backgrounds. This suggests that viewers can appreciate art independently of an artist’s personal history or reputation.
– Olive Wharry’s Artistic Talent: Olive Wharry’s artwork emerged as the favourite when respondents were unaware of the artists’ identities. This indicates that Wharry’s artistic skill and the visual qualities of her artwork resonated strongly with viewers.
Hitler’s Art and Historical Context: Adolf Hitler’s artwork garnered significant interest even before the artist’s identity was revealed, likely due to the historical context and curiosity surrounding the artwork of such a notorious figure. This curiosity may not necessarily reflect approval of Hitler’s actions but rather a fascination with his art from a historical perspective.
Minimal Impact of Artist’s Criminal Record: When informed of the artists’ criminal records, very few respondents wanted to change their image preference. This suggests that, in some cases, the public’s perception of art remains separate from their judgment of an artist’s character or actions, when art is initially presented anonymously.
Diverse Art Preferences: The varying preferences of respondents even when unaware of the artists’ identities highlight the diverse nature of art appreciation. Different individuals connect with different styles, subjects, and artistic expressions.
Complex Relationship Between Art and Artist: The research underscores the intricate relationship between art and the artist’s identity. While some individuals may separate the two, others may find it challenging to dissociate an artist’s personal history from their art. However this study has shown that artists’ criminality or notoriety had minimal affect in swaying respondents from their original art choices.
These findings emphasise the multifaceted nature of art perception and the dynamic interplay between artistic expression and the artist’s background. It’s essential to consider these complexities when discussing the influence of artists’ biographies on art reception and appreciation. It is even more important to discuss whether initially presenting an image anonymously could prove beneficial in mitigating preconceived and negative bias ahead of revealing artist information. Additionally, the results suggest that art can have a powerful and independent impact on viewers, transcending biases associated with an artist’s past if that art is presented anonymously in the first instance.
3.5 Conclusions from Poll #2
The conclusions drawn from this study highlight a consistent trend among respondents: once they’ve made their image choice, the origin of the art seems to have minimal influence on their decision, even in the case of artwork produced by someone as notorious as Adolf Hitler. Consequently, the findings from poll #2 suggest that the initial presentation of AI-generated artworks with anonymity, followed by a gradual revelation of the artist’s identity, could be an effective strategy. This approach allows viewers to establish a connection with the art based on anonymity, and this initial affinity appears to persist even after learning the artist’s identity.
3.51 Future Research:
Further studies could explore the impact of specific artist backgrounds on art preferences. For example, researching whether certain criminal activities or notoriety have a more significant influence on art perception than others could provide valuable insights into this complex relationship.
PART TWO
4. Empathy: The Role of Humanoid Robots
The second component of this study delves into the potential of empathy in reshaping perceptions of AI-generated art when anonymity is not feasible or ethical. This hypothesis was inspired by the remarkable social media engagement generated by Sun Yuan and Peng Yu’s artwork, “I Can’t Help Myself” (Preliminary Appendix), which cultivated empathy among viewers. Drawing from the findings on anonymity in the previous section, a further comparative study was designed using Facebook posts containing AI-generated art. Respondents were presented with both AI-generated art (Poll #3) and AI-generated art attributed to the fictitious humanoid robot “Athena.” (Poll #4). They were encouraged to express their perceptions.
4.1 Methodology
A comparative Study with Emotional Influence and Artist Background experiment using two further image polls (Appendix III and IV) was designed to investigate the role of empathy in art perception. Respondents were asked to rate both AI-generated artworks and those created by a “humanoid robot” named Athena, which was imbued with a poignant backstory. The primary aim was to explore the influence of information about the background of AI-generated art, specifically examining whether it could alter respondents’ perceptions and engagement..