Unit Two – Learning Objective 2
Learning Objective Two
Articulate a thorough understanding of your research and establish an informed critical position. (AC Communication)
Background
At the very start of my research paper journey, I spent a large amount of time simply trying to craft a research paper. It was important to me at this stage to get this right. I spent hours internally bouncing ideas backwards and forwards and at one point, I even enlisted the help of ChatGPT.
Once I was satisfied that I had something to work with, I decided to take a systematic approach by breaking each stage of the research down into logical stages. I also decided to undertake a research oriented version of my 21 day challenge but in a less rigid format. For 21 days I launched into a program of solid research. The end result was a completed research paper together with my thoughts, plus a behind the scenes (see end of blogpost) look at everything that went into the making of it.
Here is a link to my research paper and below are a few questions which I hope, when answered, will give a good overview of what I discovered, the arguments I formulated and what my position is regarding it all.
- What is my topic and why did I choose it?
- How and why did I choose to research it in the way I did?
- What were the key findings, implications and where do I stand on this?
- Has this impacted on my work and if so how?
- How might the research affect my work going forwards?
AI Art Negative Bias Mitigation Strategies
1). What is my topic and why did I choose it?
After extensive brainstorming, it seemed logical to research something AI related, as this is the current nature of my work. I strongly suspect that AI image generation and AI animation tools are here to stay. Therefore, knowing as much about this topic as possible seems to make sense. The title of my research paper is ‘AI Art Reception: Investigating Initial Anonymity and Empathy as Strategies for Negative Bias Reduction’. I think that this title actually sounds quite scary, but in plain English it simply means: if people don’t like AI art (Negative Bias), can the act of not telling them it’s actually AI art (Anonymity) help? And then if that doesn’t work, if we feel sorry or feel pity for the AI art in some way (Empathy), will that help us to appreciate it more?
I already had strong suspicions that AI art isn’t always favourably perceived, so for someone who has chosen to work in this area, knowing how to reduce the general feelings of negativity towards this art form is always useful. Therefore the main motivation for choosing my topic was quite simple. I like to work with AI art, unfortunately its reception isn’t always favourable – so what, if anything, can be done about it?
2). How and why did I choose to research it in the way I did?
We all consume art in a number of different ways. However, perhaps one of the most readily accessible and popular places is online. Therefore I muted the idea of conducting some primary research using image polls on the Facebook social media site. After finding a number of peer reviewed journals that supported this thinking, I felt more confident that this was both a useful and legitimate approach. Lack of time and resources meant it would be difficult to erase all of the study’s potential biases. So, I noted that this was purely a pilot study and provided a comprehensive list of its limitations.
In the first instance I decided to run an image poll among friends on Facebook. I wanted to test how they would react to images created by less conventional artists. The respondents were initially told nothing about the creators of the images (Anonymity). The images included work from: text to image AI art, AI art created by a humanoid robot called AI-DA, a work by Jackson Pollock and finally an abstract piece by a painting pig called Pigcasso. All of these polls and their findings can be found in the appendices section of the research paper – this study can be found in Appendix I. However, in brief, the most popular image was by Pigcasso the pig.
I then considered if revealing the image creators would cause respondents to change their minds and revise their choices. I was expecting respondents to move from non human creators to the human creator Jackson Pollock. This did not happen. Few, if any of the respondents wanted to change – see Appendix I. I felt this was interesting. It appeared that respondents became ‘wedded’ to their choices even when the creator was non- human.
I felt this warranted further investigation because the initial findings suggested that if someone liked a piece of art but didn’t know who initially created it (Anonymity), then perhaps they would still like it (Negative bias reduction) on learning of its AI art origins?
This moved me to create another poll, only this time using creators with varying degrees of criminality in their history. I felt that surely no-one would want a piece of art especially if it had been created by say Adolf Hitler (one of the most evil men to have ever lived). Could the revelation of criminality cause respondents to revise their image choices?
Personally I found the results startling! Four anonymous images were presented to respondents, please see Appendix II for the full study. Among the images was a painting by Adolf Hitler which was one of the more popular choices. On revealing this fact respondents were once again invited to revise their choices. I had expected that many who had inadvertently chosen Hitler’s painting would immediately want to switch to another. Interestingly, this did not happen!! Respondents on the whole remained with their initial choices. The implications for the presentation of AI art, to me, became clear. It needed to be presented in the first instance anonymously, then its creator subsequently revealed. This way an emotional connection to the art would have been already established and rendered unshakeable (evidenced by polls one and two above).
I acknowledged that it wasn’t always feasible, desirable or even ethical to present AI art in that way. Therefore, I introduced a second part to the paper to examine whether or not empathy towards an AI art creating robot might help to mitigate some of the negative bias surrounding AI art.
The thinking and inspiration behind empathy towards a robot came from a work of art called ‘I Can’t Help Myself’ by Sun Yuan and Peng Yu. Here people felt sorry for a robot that had to keep sweeping up its own hydraulic fluid that was leaking. It needed to do this in order to stay alive. I was astounded by the comments made on the Facebook post where I first discovered this piece of art. People were reduced to tears, felt emotional and likened the piece to the human condition. Clearly there was merit (strong Facebook engagement), in evoking emotions and generating sympathy. I wondered if this could be transferred to a humanoid art creating robot? In short, would people favour AI art if it came from a humanoid robot with a sad story, as opposed to art they knew was simply AI text to image generation?
I ran two further image polls. The first one was with art that respondents knew was AI text to image generated. The second was with art that allegedly came from Athena the robot who was under threat of ‘switch off’ from her evil human masters if she didn’t keep producing art.
The purpose behind the first poll was to establish respondents’ thoughts about the AI generated art and to identify any negative bias towards it. The second poll was to see if Athena the robot’s art was more favourably received through the use of empathy in mitigating negative bias. The results are discussed next but can also be found in Appendix IV.
3). What were the key findings, implications and where do I stand on this?
I ran four image polls in total. The key findings from these image polls are as follows:
* Anonymity – polls one and two in my research paper clearly showed that if artworks are presented anonymously in the first instance, an emotional connection to the work is formed. This emotional connection to the image or artwork seems to transcend the importance of the creator. Image poll two particularly highlighted this, when respondents refused to revise their image choices, even though they knew Hitler created their chosen image!! On transferring this logic to AI generated art, the initial studies suggest that anonymously presented AI art followed by creator revelation, would fare better than art presented as AI art from the start. Thus, this research suggests that initial anonymity, might in the first instance be a useful strategy in reducing negative bias towards an AI generated art piece.
* Empathy – I acknowledge in the paper that it might not always be feasible, desirable or even ethical to present something anonymously in the initial stages. There is even some research to suggest that the omission of creator details might actually negatively impact the value of an artwork. For this reason I needed to explore an additional strategy which appears in Part 2 of the research paper. After running two more image polls (see Appendix III and IV), the research indicated that, sure enough AI generated art is much better received when it comes from a humanoid robot with a sad story (Empathy). Interestingly, the first of the two polls confirmed the negative bias that I had identified via a number of research papers.
In short the key findings were as follows:
1). If you are going to present AI generated art, then in the first instance present it anonymously – allow viewers to form their own opinions and emotional attachments to it – before revealing its AI art origins.
2). When initial anonymity is neither feasible, desirable or ethical, then the findings suggest trying to evoke an air of sadness around the AI artwork definitely has a positive impact. How this might be achieved is up to the person presenting the work. However, in this study a humanoid robot called Athena with a sad story was used to invoke sympathy and empathy.
3). Hitler should have stuck to being an artist!
4). People love expressing their opinions about art through the medium of online polls.
My position on this:
Based on my findings, I believe that until people fully understand the process behind AI art, and acknowledge that it isn’t simply a clear cut case of ‘soulless plagiarism’; then AI artists working in this area should consider the findings from this short pilot study. I believe much more research is needed in this area as this study had many limitations (listed in the study) and was also very nuanced. However, it did highlight negative bias towards AI Art and also that initial anonymity and empathy could work as mitigating strategies.
4). Has this impacted on my work and if so how?
My research has definitely impacted on my work. It completely changed the way in which I approached the end of my project which culminated in a 3 minute video. In this video I used a mix of AI art, physical art, digital art and animation. However, I presented this art via Athena the humanoid robot from my research paper. I am currently awaiting peer review feedback on my video. However, I would hope that the reception of this part AI art produced video, would be better than if it was just presented as purely AI generated.
Sad Robot
Please see below the video I produced which was directly informed by my research paper.
5). How might the research affect my work going forwards?
Having conducted some initial research into negative bias mitigation strategies for AI generated art, it is very clear to me that significant negative bias towards AI generated art exists. Therefore, moving forwards I will always take this into consideration when presenting any genre of AI generated art. It may be that I would not mention the use of AI and/or use some form of empathetic narrative. However, the simple act of recognising this negative bias has already significantly changed my approach and will continue to do so moving forwards. I feel that I have come full circle in my desire to produce a human, AI and digital art hybrid (see my study statement). This is because I believe that in the absence of anonymity and empathy, probably little will improve AI art reception other than incorporating a human touch!
Recent Posts
Categories
3d 21 Day Challenge Abstract Art ai Animation Artists Augmented Reality Books Collaborations Culture Disasters Drawings Evaluations Exhibitions Experimentation Film Production Final Major Project Final Outcomes Food for Thought Galleries Ideas Lectures Little Wins Low Residency Materials Mentions My Learning my work opportunities Patterns Photography Polls and Surveys Problem Solving Reflection Research Research Paper Journey Risk Taking threats Timelapse Unit 2 Blogs Unit 3 Blogs Videos Workshops Zentangles Zoom Meetings
Tags
3d 21 Day Challenge Abstract Art ai Animation Artists Augmented Reality Books Collaborations Culture Disasters Drawings Evaluations Exhibitions Experimentation Film Production Final Major Project Final Outcomes Food for Thought Galleries Ideas Lectures Little Wins Low Residency Materials Mentions My Learning my work opportunities Patterns Photography Polls and Surveys Problem Solving Reflection Research Research Paper Journey Risk Taking threats Timelapse Unit 2 Blogs Unit 3 Blogs Videos Workshops Zentangles Zoom Meetings