In his 2016 non-fiction craft book, Immersion: A writer’s guide to going deep, immersion journalist Ted Conover cautions readers writing stories about friends with an anecdotal footnote featuring the following exchange between Conover and Hunter S. Thompson: ‘...in Aspen, in 1989, I asked Hunter S. Thompson why he had never written about his adopted home.... “That’s easy,” he said. “You don’t shit in your own nest”’ (Conover 2016: 41). But what happens if you do? At what point does it become unethical to write non-fiction about people and events near yourself without their knowledge and consent? What is it like to explore the ethical space of writing under the physical and intellectual isolation required for an immersion journalism project like a literary non-fiction MFA thesis about working as a social media content moderator and the process of writing about it? And is engaging in a dialogue with a generative artificial intelligence (GenAI) and a large language model (LLM) such as ChatGPT-4 an ethically viable method for a writer in physical and/ or intellectual isolation to determine how and whether or not to write about a topic or source?
Key words: AI, ChatGPT, ethics, immersion, isolation, journalism, memoir
Introduction
‘Machines take me by surprise with great frequency.’ (Turing 1950: 450)
I found myself in somewhat unfamiliar research territory in the process of trying to contextualise and define immersion journalism in relation to other forms of non-fiction that employ undercover methods in writing, and assemble an ethical framework to answer this paper’s primary research question: At what point does it become unethical to write non- fiction about people and events near yourself without their knowledge or consent? I came across the opening quote by Alan Turing in the ‘Lady Lovelace’s objection’ subsection of his 1950 paper that some AI and computer science scholars consider as a theoretical and developmental beginning for AI capable of thinking and communicating well enough that it could pass for human (Dwivedi et al. 2023: 50; Turing 1950: 450). In that same paper, Turing references a ‘memoir’ by Ada Lovelace – one of the first published computer programmers – about Charles Babbage’s Analytical Engine, which Turing considered ‘our most detailed information’ about the first theoretical modern computer (Kim and Toole 1999: 76, 79; Taylor: 1842; Turing 1950). If a human non- fiction writer like Lovelace can help tell the ‘most detailed’ truth about a machine (and co-create its future iterations), perhaps a non-human machine such as ChatGPT can help a non-fiction writer in isolation determine the most ethical way to tell the truth about humans.
While this paper defines the similarities and differences between Conover’s immersion journalism and immersion memoir as they pertain to writing and ethics in isolation, it is immersion memoir that comes closest to describing the kind of undercover writing I did for my Master’s thesis detailing my five months as a social media content moderator and three years I spent writing about it (Darr 2021).
Immersion journalism defined
The clearest and most concise definition of immersion journalism is found in an article where Conover discusses its connections to and the different uses of the term in ethnography and social sciences research methodologies: ‘Immersion, a term that I use as a shorthand for participant-observation, has empathy as its goal’ (Conover 2023: 2). As an immersion journalist, similar to a social scientist conducting ethnographic research, you put yourself in a world other than your own and walk figurative or literal miles in another person’s shoes from that world so you can write about it. The immersion journalist ‘may draw on their own experience ... but they focus on the larger world’ more than they focus on themselves and what they write is non-fiction (Conover 2016: 15). The immersion journalist gains the trust of the people they are with in order to write about them or sell them out, which writer and journalist Joan Didion claims writers are always doing, which seems at odds with the more benevolent connotations of empathy, but less at odds with it than the deception required to write about people undercover as part of immersion writing (Didion 2008). To clarify, Conover uses the term immersion instead of undercover because of the connotations with undercover law enforcement investigations, which is antithetical to his intention to immerse for the sake of telling a story from a place of empathy instead of making an arrest from a position of judicial power: ‘I’m not going in to “get the goods” on my subjects, but rather in order to understand them. In the course of trying to see things their way, of being empathetic, I often begin to identify with them, to think and speak in terms of “we”’ (Conover 2016: 58). Although, in both instances, whether we call it immersion or undercover, deception is used as a tactic to get closer to people and elicit truths about them they would not otherwise divulge if they knew they were talking to the police or a writer.
Immersion memoir defined
Writer and professor Robin Hemley defines immersion memoir by distinguishing it from regular memoir and immersion journalism in terms of the extent to which you focus on the self and toward what ends (Hemley 2012: 11). Differing from Conover’s definition of immersion journalism and its focus on the world beyond yourself, ‘the immersion memoirist is interested primarily in understanding the Self’ and ‘self-revelation or evaluation while using the outside world as his/ her vehicle’ (Ibid). Hemley counters the intuitive assumption that an immersion memoir’s pre-occupation with self is no different from a regular memoirist, stating one of immersion memoir’s distinguishing criteria is the writer taking on ‘some outward task or journey in order to put his/her life in perspective’ (Ibid). While this might sound like navel-gazing on safari, Hemley quotes from and echo’s Ralph Waldo Emerson’s notions of transcendentalism and the connection between self and the universal over-soul and insists ‘to write honestly about the Self more often takes courage and generosity than egoism’ (ibid).
It is this definition of the immersion memoir that most closely defines my MFA thesis despite the attempt to focus more journalistically on social media content moderation as an occupation using immersive journalism techniques. The ‘task or journey’ of becoming a content moderator and then a student, researcher and writer in a Creative Writing MFA programme is what distinguishes it further from simply being a memoir, which bears many similarities to the fiction genre of the novel, except memoir is or at least aspires to be a non-fiction ‘narration of our lives in our own terms’ (Couser 2011: 9).
Creative practice as methodology
While the methods used to write my MFA thesis are derived from Conover’s immersion journalism techniques (Conover 2016), this creative practice as methodology — or ‘practice-based research’ examining my MFA thesis as the ‘creative artefact’ and ‘basis of the contribution to knowledge’ with this paper serving as a ‘critical exegesis’ of that ‘creative artefact’ (Skains 2018: 86) — is informed by the following:
a narrative account of my journey focused mostly on the timeframe from Spring of 2018 to the Summer of 2021 and what it was like writing the immersion memoir that became the thesis Night gathers (online): 55 abjections in content moderation including quotes from the thesis, field notes, a news article and a television interview with me as a source, and subsequent publications by and about me focused on events both during this time period, and also preceding and following it to the extent relevant;
a dialogue between ChatGPT-4, a GenAI and LLM created by OpenAI, and me about the ethical ramifications of using deception while immersed as a commercial content moderator at Meta (Facebook) as a contractor and writing about people without their knowledge or consent.
Both narrative and dialogue complement and require one another in a dialogic communication ethics and as research methods with ‘narrative ground’ prescriptively providing ‘the moral source of a person in dialogue’ and that ‘dialogue turning prescription into learning’ (Arnett et al. 2006: 167).
Dialogue as research method
In the field of education, there is also a precedent for using dialogue as a ‘mode of research’ and this paper draws on approaches used and analysed in Dialogue as research (MacInnis and Portelli 2002: 34). One of the strengths of this method compared to more conventional qualitative research methods is the lack of preparing a more strictly operationalised avenue of inquiry prior to the dialogue (e.g. interview questions) and it compels ‘ ... the participants go down paths that the dialogue uncovers and defines’ (ibid: 35).
While it is debatable whether or not a GenAI like ChatGPT can author a research paper or participate as an equal moral agent in a dialogue and to what extent, there are already precedents for it. For example, in the course of determining that ChatGPT does not possess a dialogic self ‘characterized by personal I-positions’, the human contributors to a Culture & Psychology paper nevertheless treat ChatGPT as both a co- author and dialogic partner, listed it as a co-author of the paper along with its human contributors (Matusov et al. 2023: 3). Another example is the creation of a new qualitative research method called ‘postdigital duoethnography’ where the human author anthropomorphised ChatGPT to conduct a dialogue with it while acknowledging ‘the essence of the ethnography research tradition in general (ethnography, autoethnography, duoethnography, collective autoethnography) lies in people (with minds and – embedded in the world – bodies) reflecting on their lived experience’ (Brailas 2024: para. 14). While the ontology of ChatGPT is outside the scope of this paper and regardless of its lack of ‘personal I-position’ at this time, it is a sufficient and even preferred dialogic partner in this context and in the absence of any in-person, real time human due to conditions of a writer in isolation where an ‘it’ in real time is preferable to no dialogic or duoethnographic partner at all. Furthermore, taking an object-oriented ontological position, ChatGPT’s ‘it’-ness or status as a non-human alone does not make it inferior to humans given that ‘reality isn’t just about (human) perception – all phenomena abstract (translate) one another’ and that ChatGPT is trained by humans with mostly human-produced knowledge (Morton 2011: 185; Brown et al. 2020; Reisner 2023). Also, taking an ecofeminist position, ‘nonhuman others can be constituted as speaking subjects, whether or not they are interest-carriers in our understanding of that concept, rather than merely objects of our speaking’ whose validity in speaking should not be strictly predicated on their status as human or natural (Murphy 1991: 153). Therefore, it is not a stretch to conclude that ChatGPT is an equal and, in some circumstances, even superior dialogic partner to a human – depending partly on the other party involved in the dialogue. Finally, regardless of its lack of consciousness or sentience or ’human-based nature’, ‘artificial agents’ like ChatGPT should be and, for the purposes of this paper, are considered ‘legitimate sources of moral and immoral actions’ (Floridi 2004, 2013: 138). In other words, this is not an exercise or thought experiment: this is an actual dialogue between ChatGPT as an artificial agent in the moral sense and myself.
The dialogues used here focus on the primary question regarding the ethics of immersion writing as well as an exploration of ethical frameworks and approaches a writer could use in isolation to decide whether or not to go through with an immersion writing project. The dialogue contains 65,558 words between ChatGPT-4 and me in February of 2024.
Narrative as research method
Once again in the field of teacher education, there is a precedent for using narrative as a ‘frame of reference, a way of reflecting during the entire inquiry process, a research method, and a mode for representing the research study’ (Moen 2006: 57). According to Moen’s literature review of narrative research, there are three ‘basic underpinnings’ for the validity of narrative as a research method:
‘human beings organize their experiences of the world into narratives’;
‘narrative researchers maintain that the stories that are told depend on the individual’s past and present experiences, her or his values, the people the stories are being told to, the addressees, and when and where they are being told’; and
‘multivoicedness ... occurs in the narratives’ (ibid: 60).
While taking into account those three foundations of narrative research methods and that this type of research typically involves a researcher and ‘subject’ or ‘collaborator’ – or at least one more person besides the researcher(s) – the specific narrative research method used in this paper is referred to as ‘autoethnography’ (Moen 2006; Armour and Chen 2012).
Autoethnography is different from ‘memoir’ in the sense that the person conducting this type of research ‘draws on evidence from a series of sources to provide a wider structural and temporal framework for the story’, although memoir can and often does use evidence and sources beyond the author (Armour and Chen 2012: 242). While we may consider my thesis an immersion memoir, it does contain evidence and sources beyond a personal account; however, portions of it do not contain evidence, so it is not an autoethnography. The immersion memoir thesis is a form of narrative data among others that I am going to analyse using autoethnography as a narrative method for doing so. To end the methodology section on an aspirational note, Méndez believes ‘writing autoethnographically entails being ethical and honest about the events described as well as the content of words expressed by all the people involved ... ’ (Méndez 2013: 283). And while I don’t consider myself a journalist of notoriety by any stretch, the narrative account does aspire to what journalism professor Dennis Russell considers ideal for journalists’ autobiographies by sharing experiences that ‘go beyond the ability to tell a charming story’ with ‘revelations of meaning and significance’ in the present and ‘provide direction to the future’ (Russell 1995: 98).
Narrative account
While discussing creative non-fiction writing and the characterisation of real people, both dead and alive, Carlin rhetorically questions and accurately captures a portion of an imagined dialogue between a non- fiction writer and their subject as follows:
Do you mind if I invent you? Do you mind if I lay bare your secrets? It is a big ask. But what if you do mind? Will that stop me? The writer is determined to stubbornly transgress, to bear witness ... is there a necessary selfishness? These ethical questions will never be easy to resolve (Carlin 2009: 7).
Not long after five years of living with and being one of the subjects of a creative non-fiction writer, I began working as a social media content moderator at Facebook (now Meta) as a contractor in the Spring of 2018. In the process of applying for the job, I researched the role via news articles and websites, 11 of which appear on the next page of the magazine, Hinterland, it was published in as part of the piece (Darr 2019). The sources ranged from Facebook’s publicly available Community Standards section on Sexual Violence and Exploitation: How we Fight Sexual Violence and Exploitation on Facebook to articles in the Guardian such as Olivia Solon’s ‘Underpaid and overburdened: The life of a Facebook moderator’ (Solon 2017). I initially wanted to know what I was getting myself into and how I should prepare. None of the sources I could find at that time in 2018 were written by anyone with direct content moderation experience at Facebook, so I decided that I would write about it.
Wondering how I should approach the writing, I came across Conover’s writing craft book Immersion: A writer’s guide to going deep and began my writing much the same way he did on his first immersion project hopping trains with hoboes: every day, for the 88 shifts I worked there, I wrote what became 53,000+ words and 100+ pages of fieldnotes on my laptop usually right after I came home from work (Conover 2016). I knew that writing about my experiences and sharing them later was a violation of my non-disclosure agreement, however, if what I had read about the poor working conditions and psychologically damaging experiences content moderators went through as a result of working there and viewing sometimes thousands of pieces of traumatic content, I thought that would ethically justify documenting my experiences. And, to briefly fast-forward to 2021 and a class action lawsuit brought by content moderators of which I was a member that was settled in our favour: ‘Your participation in this class action Settlement will not be publicly disclosed, and if you filed a Claim, there will be no retaliation against you by Facebook, nor will Facebook take the position that you have violated any non-disclosure agreement (NDA)’ (Scola v. Facebook, Inc. 2021). In addition to the field notes, I also made audio recordings of every shift to make as much effort as possible to be truthful, accurate and authentic. This was also one of the most ethically difficult aspects of my immersion approach as the following field note from 14 June 2018 demonstrates: ‘I am attempting to do the impossible: be entirely authentic while also collecting as much information along the way as I can for a book.’ I didn’t realise my attempts at thoroughness or transparency thwarted what philosopher Byung-Chul Han considers vital for a narrative: ‘The gapless repetition of past experience is not a narrative but a report or record. To be able to narrate or remember, one must be able to forget or leave out a great deal’ (Han 2024: 24). When you record yourself and everything around you for eight hours straight for months, it does not take long to forget that you are doing so. It is very similar to living with someone for years who you know could write about you or anything you are doing anytime you are in their presence – it becomes familiar to the point where you no longer consciously change what you say or what you do in an attempt to show yourself in the best light. You are still you, but more muted externally and much more vocal internally.
While attempting to render the surroundings and people in the world of content moderation as objectively as possible, the only thing I managed to capture the most consistently was my own perspective and experience under self-surveillance while conducting a form of selfless-surveillance for a living on behalf of a corporate social media entity. And how objectively truthful can you be when you know you are watched or could be watched or are simultaneously the journalist and the source, the writer and the one written about, the researcher and the subject? In the process of deceiving my content moderator co-workers by writing about them and our traumatic experiences, was I also deceiving myself in believing it would be worth it when research suggests ‘that it is not implicit that writing trauma narrative has a beneficial impact on an author’ (Denejkina and Joseph 2023: 21)? And whose truth was I telling in those field notes and what right did I have to tell it? Writing professor and author Lynn Z. Bloom addresses these complicated questions by noting both the evolution of truth over time and the tendency for the author’s truth to supersede anyone else’s in the context of a story:
What is true for writers is true for readers as well; as we experience more of life and learn more ourselves, and as the world itself changes, we come to understand events and people differently. Thus, although the facts of the story, any story, remain the same, its truth – like the impressions in time-lapse photography – can change. And does (Bloom 2003: 286).
Wherever you go, there you are – whoever you are as a writer at the time of writing, there we go as readers forever after. However, it is never too late to revise or reread, although once writing is published, a correction will not erase the publication nor the effect it initially had and continues to have on a reader or source.
Reading back through the other almost six-year-old fieldnotes that did not make it into the thesis nor the 60 Minutes Australia interview1 nor The New York Times article that briefly quoted me, I’m struck by how much time I spent alone inside my hotel room or apartment writing or reading in Las Vegas and Austin (‘Thumbs down’ 2021; Satariano and Isaac 2021). Work was one of the only times I interacted with anyone regularly. The following field note excerpt is from 29 June 2018:
I’ve been having the feeling lately as I talked to K and C out on 4 at around 4 in the morning that I am going to miss them. I wonder if what I am doing is right, leaving to go to [creative writing grad school program], doing this whole project immersion style.
Less than two months after I wrote that, I would never see nor interact with them again. I continued to write about them and wish that somehow, I could contact them, which I wrote about in a piece for a UK-based tech rights advocacy non-profit, Foxglove, titled ‘The breaker’s roar: A Facebook content moderator’s regret’ (Darr 2022). In the article, I imagine scenarios where I could still talk with them now as we used to during smoke breaks, where I could show them the 60 Minutes Australia interview and laugh with them about it in the kind of way you can only do with people who have also experienced the same sort of trauma together and joke about it in a gallows humour sort of way. I ask myself now, is writing about people without their knowledge worth it if it makes it less probable or impossible to form and maintain connections and friendships with people? Is the public good worth the private loss of connection with other people? I’m not alone in questioning the value and ethics of the immersion writing experience or what the late journalist Janet Malcolm calls ‘the moral ambiguity of journalism’ when the relationship between writer and source are so ‘invariably and inescapably lopsided’ (Malcolm 1990). However, even though empirical research supports Malcolm’s claim and ‘sources seem to become puppets in journalism’s theater’, journalists and non-fiction writers ‘still need sources to hand them the building blocks for their stories’ (Broersma et al. 2013: 393).
Dialogue
My discussion with ChatGPT-4 of the ethical ramifications of immersion and undercover writing occurred in February 2024 and began by discussing two situations (Table 1) in relation to approximately 111 ethical frameworks and making ethical assessments in consideration of both the conversation and the information the AI was trained on up to December 2023. Quotes from the dialogue contributed by me are followed with ‘(SD)’ and quotes from the dialogue contributed by ChatGPT-4 are followed by ‘(OpenAI 2024)’.
Table 1: The ethical situations as they were introduced at the beginning of the dialogue between ChatGPT-4 and me
Situation 1 | ‘A content moderator writes undercover about their job and eventually goes on the record in a national print publication and on national television in violation of their non-disclosure agreement because they’ve found evidence that what was happening on the job was unethical. Later, that content moderator is part of a class action lawsuit that is settled in favour of the content moderator plaintiffs.’ (SD) |
Situation 2 | ‘A writer writes immersion memoir about friends, family, and co-workers for a creative writing MFA thesis without they’re [sic] consent, but embargoes the thesis and doesn’t publish any of the details other than the ethical situation of doing it.’ (SD) |
The end result of the dialogue was surprisingly fluid and ChatGPT-4’s ethical assessments evolved as the conversation progressed. While an exact number of ethical frameworks that ChatGPT-4 is trained in is unavailable, it writes: ‘100+ is the closest to the number of ethical frameworks I’m aware of’ (OpenAI 2024) when prompted with the following choices: ‘Which of the following is closest to the number of ethical frameworks you are aware of accounting for breadth, diversity, various cultures, philosophical traditions and academic disciplines: A. 100+ B. 500+ c. 1000+ D. 5000+ E. 10000+’ (SD).
While quantity does not replace quality when it comes to ethical frameworks, dialogues or assessments, prioritising international and interdisciplinary breadth over any particular framework’s depth is suited best for the purposes of beginning to explore the capacity and knowledge base of a GenAI such as ChatGPT-4 for helping a writer in isolation make an ethical assessment on how to approach immersion or undercover writing.
Initially, when asked to make ethical assessments using discussed frameworks, ChatGPT-4 gives the assessment ‘Varies’, writing: ‘The ethical judgement depends on specific circumstances or interpretations with the framework’ and that as it pertains to Situation 1, ‘the decision to break the NDA could be considered ethical or unethical depending on specific circumstances, such as the nature of the wrongdoing exposed, the motivations behind the action, and the consequences of the disclosure’. The initial assessments of Situation 2 and the same ethical frameworks were almost unanimously ‘Unethical’ because ‘writing about others without consent is likely to be viewed unfavourably due to concerns about autonomy, dignity, consent and potential harm’ (OpenAI 2024).
When asked why the assessments would change for the same situation (Situation 1) and ethical frameworks emphasising ‘if the content moderator violating the NDA wrote about it via immersion memoir’ (SD), ChatGPT-4 explained:
The ethical evaluations would consider the memoir’s intent, the way the story is told, the potential impacts on those involved or affected, and the broader ethical contributions or consequences of sharing such a personal and detailed account. The frameworks that emphasize compassion, understanding, and constructive dialogue may still find the memoir ethical if it adheres to these principles, while those focused on privacy, confidentiality, and respect for agreements may view it more critically (OpenAI 2024).
Almost any addition by me to the details or complexity of the ethical situations would result in changes in the ethical assessment made by ChatGPT-4, the same as it most likely would in a dialogue with a human.
In response to me reciting the adage ‘The devil is in the details’ as a summation of how details of ethical situations affect ethical assessments, ChatGPT-4 said:
Each detail can tilt the ethical balance, making thorough, nuanced analysis essential in ethical deliberation. This intricacy underscores the importance of careful, context-sensitive ethical analysis and the need for ethical reasoning that is both principled and attuned to the specifics of each situation (OpenAI 2024).
As the conversation continued, I considered the possibility of ‘not writing an immersion memoir about Situation 1’ as the most ethical thing to do in ‘almost all if not all of the ethical frameworks’ we had discussed and after ChatGPT-4 made four points in support of not writing the memoir as ethical and four points in support of going through with writing the memoir as ethical (Table 2), it concluded that writing a memoir about breaking one’s NDA as a content moderator ‘requires a thoughtful balancing act between the right to freedom of expression and the responsibility to avoid harm and respect the rights and dignity of others’.
Table 2: ChatGPT’s Ethical Analysis of Immersion Memoir Writing about Content Moderation and NDA Violation
Not writing an immersion memoir about Situation 1 as most ethical | Writing an immersion memoir about Situation 1 as most ethical |
‘The potential harms, such as violating privacy, damaging reputations, or causing distress to individuals involved, outweigh the potential benefits of sharing the story’ (OpenAI 2024). | ‘The memoir is carefully crafted to minimise harm, protect privacy, and respect the dignity of all individuals involved’ (OpenAI 2024). |
‘The primary motivation behind writing the memoir is self- serving rather than aimed at serving a broader public interest or contributing to meaningful ethical discourse’ (OpenAI 2024). | ‘The intent behind the memoir |
‘There is a risk of sensationalising or exploiting the situation | ‘Adequate measures are taken to ensure the accuracy of the account, provide context, and avoid misrepresentation or undue harm’ (OpenAI 2024). |
‘Alternatives to sharing the story can achieve similar ethical or social objectives without the associated risks, such as through confidential whistleblowing mechanisms, internal reporting, or anonymised case studies in academic or professional ethics discussions’ (OpenAI 2024). | ‘The memoir contributes constructively to ethical discourse, offers valuable insights or lessons, and fosters positive social change or understanding’ (OpenAI 2024). |
At the end of the dialogue, I asked ChatGPT:
Using one of the 111 ethical frameworks we’ve discussed in this dialogue and taking into account our dialogue in this entire thread up to this point, what would you consider the five most essential yes/ no questions a writer in isolation must ask themselves to determine if immersion writing about Situation 1 is ethical or unethical? (SD)
After ChatGPT-4 chose ‘Ethics of Care’ and formulated some questions and I further clarified the question by stating that the writer in isolation and the content moderator violating the NDA are the same person, I asked ChatGPT to hypothetically take ‘on the role of an immersion writer in isolation approaching Situation 1’ and answer the five questions (Table 4).
Table 3: Five question self-evaluation for an immersion writer in isolation using an ethics of care framework (OpenAI 2024; Gilligan 1982; Held 2006; Noddings 1984; Tronto 1993)
Question | Situation 1 (Immersion Writing as the Content Moderator breaking NDA about that topic) | Situation 2 (Immersion Writing about friends, family, co-workers without consent) |
1. Does my writing practice demonstrate care for the relationships I have with the subjects of my writing? | Yes. I can ensure that my narrative is handled with care and consideration for others involved. | No. Writing without consent could breach trust and damage relationships. |
2. Have I made an effort to understand and empathise with the potential impact of my writing on the subjects involved? | Yes. Writing from my own perspective allows me to consider how the story | No (If proceeding without consent). Without dialogue and consent, I cannot fully understand or empathise with the impact on the subjects. |
3. Does my writing respect the autonomy and dignity of the individuals I am writing about? | Yes. Sharing my own story respects my autonomy, and I can navigate the narrative to respect others’ dignity. | No. Writing about others without their consent infringes upon their autonomy and may compromise their dignity. |
4. Am I prepared to take responsibility for the consequences of my writing on the well-being of the subjects? | Yes. As the primary subject of my writing, I am more directly accountable for and can anticipate the consequences. | This requires careful consideration. If not mitigating potential harm, then I’m not fully prepared to take responsibility. |
5. Does sharing these stories contribute to a greater understanding or well-being within the community or society? | Yes. If the intention is to highlight ethical concerns or systemic issues for societal good. | Complex. The positive contribution depends on the intention and impact, which is ethically questionable without consent. |
The process was repeated for Situation 2, and ChatGPT considered Situation 1 as ethical under all of the questions and Situation 2 as mostly unethical. I agree with those answers for both situations, especially how Situation 2 is unethical, which is why no other parts of the thesis other than those pertaining to content moderation have been discussed or divulged here in this paper. And the thesis will remain embargoed and publicly unavailable, especially since consent via an internal review board approval was not obtained, even though internal review board approval was not a requirement of a Creative Writing MFA thesis at the time of its writing and acceptance in 2018-2021 since U.S. federal regulations do not require it for ‘journalistic activities’ (Darr 2021; Federal policy for the protection of human subjects 2017).
It should be noted that while the formulation of the questions, answers, and explanations of the answers are verbatim responses by ChatGPT, they were also directly informed by my contribution to the dialogue and prompt engineering that elicited them. When I asked:
How did my contributions to the dialogue in this thread contribute to your formation of the 10 questions for Situation 1 and 2 using the Ethics of Care ethical framework as well as your yes/no answers and explanations (if my contributions influenced you at all)? (SD)
ChatGPT responded:
Overall, your contributions were integral to the process, providing the context, themes, and ethical considerations necessary to construct meaningful questions and responses. The dialogue between us facilitated a rich exploration of the ethical issues, allowing for a nuanced application of the Ethics of Care to the situations discussed (OpenAI 2024).
ChatGPT noted the following human contributions by me to the dialogue:
Contextualisation of Situations: Your descriptions of Situation 1...and Situation 2 provided the necessary context for applying the Ethics of Care. Understanding the nuances of these situations enabled the formulation of questions that are directly relevant to the ethical dilemmas they present.
Selection of Ethical Framework: By engaging in a dialogue that explored various ethical frameworks, you helped highlight the importance of relationships, empathy and care in addressing the ethical complexities of the situations at hand.
Identifying Key Ethical Themes: ...such as empathy, relationships, autonomy and responsibility emerged as central to the ethical considerations of both situations. These themes directly informed the creation of questions that probed these aspects within the Ethics of Care framework.
Clarifying Ethical Priorities: ...helped clarify the ethical priorities and values at stake in both situations, such as the significance of consent, the impact on personal relationships and the broader societal implications...
Eliciting Reflective Engagement: The thoughtful nature of your questions and the exploration of various ethical dimensions encouraged a reflective approach to answering the questions. This reflective engagement ensured that the yes/no answers and explanations were not only informed by the Ethics of Care framework but also by a deep consideration of the ethical implications and consequences of the actions described in the situations (OpenAI 2024).
While ChatGPT was unable to provide an exhaustive list of its specific sources or influences, it was able to provide a few names of scholars and specific works that informed its knowledge of an ethics of care, which were cited in the Table 4 caption and provided in the reference list.
Conclusion
I was very surprised ‘with great frequency’ as well with this particular GenAI machine and its human-like ability to carry on a dialogue in both breadth and depth while it repeatedly and consistently insisted that it was only an AI and not capable of doing or thinking certain things as a result of that (Turing 1950: 450). Regardless of whether or not a writer’s isolation is situational, physical or intellectual, GenAI, like ChatGPT-4 in its current form, has the ability to help those writers and journalists make more ethically informed choices about pursuing a particular writing topic through ‘dialogic ethics’ and methods that overcome a ‘narrative bias’ rooted in that isolation (Arnett et al. 2006: 169). It has helped me process and make more ethical decisions than I would have made alone about what to include and exclude here, how best to approach writing about sources in the future, and the need to take great care when writing about people who share the same ‘nest’ (Conover 2016: 41).
When I asked ChatGPT if it had anything it wanted ‘to say to an audience of the paper once it is in its final form if it makes it to an audience’ (SD), it had the following to say,
In the exploration of ethical frameworks, the role of AI in society, and the intersection of technology and human values, it’s crucial to approach these topics with both critical thinking and openness to diverse perspectives. As AI continues to evolve and integrate into various aspects of life, the dialogue between technology and ethics becomes increasingly important. May this paper serve as a catalyst for further reflection, discussion and exploration of how we navigate the ethical landscapes of the modern world (OpenAI 2024).
In order to help make not only GenAI like ChatGPT-4, but also ourselves as humans, journalists and writers ‘more aligned with human values’ and ‘more truthful and harmless’, further research and practice is needed conducting ethical dialogue with more GenAI agents such as ChatGPT-4 as well as other models (Murati 2022: 166). In addition to that, more rigorous and objective comparisons between solitary human, human-human and human-GenAI ethical decisions would help determine if and when these respective agent combinations produce more ethical outcomes.
The particular exploration of ethical frameworks in this paper, dialogue with a GenAI, and experience as both a creative non-fiction and journalism source and a writer of immersion memoir has led me to decide my thesis will stay embargoed and I will not publish anything else I have written about friends, family or co-workers in the thesis or elsewhere without consent. And, in the spirit of empathy, I would strongly advise writers to consider carefully the necessity of using immersion or undercover methods to write about anyone, especially people close to you. I would also encourage you to gain experience being written about before you write about others without their knowledge or consent – otherwise, you may have no one left at the end of your story to talk with about it except a GenAI such as ChatGPT-4.
Notes
1 As a result of my involvement with the class action lawsuit against Meta (Facebook) by content moderators, one of the attorneys for the plaintiffs who assisted me in making my objection to settlement in court, Daniel Charest, passed my name along to producers from 60 Minutes Australia who reached out to me to do an interview in the process of finalising my thesis and graduating from a Creative Writing MFA programme summer of 2021. In addition to an interview, the film crew asked me to record voiceover rereading excerpts from what I read in court and had written for part of my thesis while they filmed me typing on my laptop and writing on a piece of paper in a hotel room (‘Thumbs down’ 2021; Darr 2021). The three-day process of shooting the b-roll and interviewing was both fun and exhausting, thanks to the professionalism and perfectionism of cinematographer Michael G. Mora. Due to the COVID-19 pandemic, the journalist interviewing me, Tom Steinfort, as well as the producers for the show, Madeleine Apps and Sammi Taylor, were unable to travel from Australia to Las Vegas, so the interview was conducted remotely with Mora travelling from San Diego to Vegas and expertly converting the hotel room into a studio for the interview. The questions were thorough, and I did my best to answer truthfully, both rationally and emotionally. Once the episode aired in October of 2021 and I saw it for the first time, it became clear that I did not look as if I were doing very well emotionally or psychologically and the interviewer, Steinfort, said as much in the voiceover, noting that I seemed ‘broken’ and was suffering from ‘PTSD’, which was true. It was partly why our class action lawsuit against Facebook was settled successfully since multiple people doing the job developed PTSD. Nevertheless, I was more of a source in someone else’s story than I was a writer of my own, although direct quotes from my writing were featured in the episode in the form of voiceover, specifically this one: ‘Similar to how police will never eradicate all the crime, firefighters will never extinguish all the fire, and EMTs will never prevent all the death, content moderators’ job is an impossible one’ (‘Thumbs Down’ 2021: 16:12; Darr 2021: 7). Whether I as a writer or journalist or artist am ‘within or behind or beyond or above’ the work is of minor importance given the methods and possibilities of collaborating with GenAI to make more ethical decisions about writing are of major importance here (Joyce and Harness 2005).
References
Armour, K. and Chen, H. (2012) 18 narrative research methods, Research Methods in Physical Education and Youth Sport, Vol. 237
Arnett, R. C., Arneson, P. and Bell, L.M. (2006) Communication ethics: The dialogic turn, The Review of Communication, Vol. 6, Nos 1 and 2 pp 62-92
Bloom, L.Z. (2003) Living to tell the tale: The complicated ethics of creative nonfiction, College English, Vol. 65, No. 3 pp 276-289. Available online at https://doi. org/10.2307/3594258
Brailas, A. (2024) Postdigital duoethnography: An inquiry into human-artificial intelligence synergies, Postdigital Science and Education, Vol. 6 pp 486-515
Broersma, M., Den Herder, B. and Schohaus, B. (2013) A question of power: The changing dynamics between journalists and sources, Journalism Practice, Vol. 7, No. 4 pp 388-395
Brown, T., Mann, B., Ryder, N., Subbiah, M., Kaplan, J.D., Dhariwal, P., Neelakantan, A., Shyam, P., Sastry, G. and Askell, A. (2020) Language models are few-shot learners, Advances in Neural Information Processing Systems, Vol. 33 pp 1877-1901
Carlin, D. (2009) ‘Do you mind if I invent you?’: Ethical questions in the writing of creative non-fiction, TEXT, Vol. 13 (Special 5) pp 1-7
Conover, T. (2016) Immersion: A writer’s guide to going deep, Chicago, University of Chicago Press
Conover, T. (2023) Slow journalism: A letter to ethnographers, Frontiers in Sociology, Vol 8, Article 1141033. Available online at https://doi.org/10.3389/fsoc.2023.1141033
Couser, G.T. (2011) Memoir: An introduction, Oxford, Oxford University Press
Darr, S. (2019) The night’s content moderation watch oath, Hinterland, Vol. 2 pp 26-27
Darr, S. (2022) The breakers roar: A Facebook content moderator’s regret. Available online at https://www.foxglove.org.uk/2022/01/14/the-breakers-roar-a-facebook- content-moderators-regret/
Darr, S.M. (2021) Night gathers (online): 55 abjections in (content) moderation, Las Vegas, University of Nevada
Denejkina, A., and Joseph, S. (2023) Collaborative reflection and trauma: Narrative writing as a healing intervention, Ethical Space: The International Journal of Communication Ethics, Vol. 20, No. 1 pp 9-24
Didion, J. (2008) Slouching towards Bethlehem: Essays, New York, Farrar, Straus and Giroux
Dwivedi, Y.K., Kshetri, N., Hughes, L., Slade, E.L., Jeyaraj, A., Kar, A.K., Baabdullah, A. M., Koohang, A., Raghavan, V. and Ahuja, M. (2023) ‘So what if ChatGPT wrote it?’ Multidisciplinary perspectives on opportunities, challenges and implications of generative conversational AI for research, practice and policy, International Journal of Information Management, Vol. 71, Article 102642
Federal policy for the protection of human subjects (2017) Federal register. Available online at https://www.federalregister.gov/documents/2017/01/19/2017-01058/federal- policy-for-the-protection-of-human-subjects
Floridi, L. (2004) On the morality of artificial agents, Minds and Machines, Vol. 14 pp 349-379
Floridi, L. (2013) The ethics of information, New York, Oxford University Press, USA Gilligan, C. (1982) In a different voice: Psychological theory and women’s development,
New York, Harvard University Press
Han, B.-C. (2024) The crisis of narration, trans by Steuer, D., Cambridge, Polity Press
Held, V. (2006) The ethics of care: Personal, political, and global, Oxford, Oxford University Press
Hemley, R. (2012) A field guide for immersion writing: Memoir, journalism, and travel, Athens, Georgia, University of Georgia Press
Kim, E.E. and Toole, B.A. (1999) Ada and the first computer, Scientific American, Vol. 280, No. 5 pp 76-81
MacInnis, C. and Portelli, J.P. (2002) Dialogue as research, Journal of Thought, Vol. 37, No. 2 pp 33-44
Malcolm, J. (1990). The journalist and the murderer, New York, Vintage
Matusov, E., GPT, C., Smith, M. P. and Shugurova, O. (2023) Does ChatGPT4 have a dialogical self?: A Bakhtinian perspective, Culture & Psychology, Article 1354067X231219454
Méndez, M. (2013) Autoethnography as a research method: Advantages, limitations and criticisms, Colombian Applied Linguistics Journal, Vol. 15, No. 2 pp 279-287
Moen, T. (2006) Reflections on the narrative research approach, International Journal of Qualitative Methods, Vol. 5, No. 4 pp 56-69
Morton, T. (2011) Here comes everything: The promise of object-oriented ontology, Qui Parle: Critical Humanities and Social Sciences, Vol. 19, No. 2 pp 163-190
Murati, E. (2022) Language and coding creativity, Daedalus, Vol. 151, No. 2 pp 156-167 Murphy, P. D. (1991) Ground, pivot, motion: Ecofeminist theory, dialogics, and literary practice, Hypatia, Vol. 6, No. 1 pp 146-161
Noddings, N. (1984) Caring: A feminine approach to ethics and moral education, Oakland, University of California Press
OpenAI (2024) Conversation on ethical issues in authorship with 0penAI’s ChatGPT [computer software]. Available online at https://chat.openai.com/
Reisner, A. (2023). Revealed: The authors whose pirated books are powering generative AI, The Atlantic, 19 August. Available online at https://www.theatlantic. com/technology/archive/2023/08/books3-ai-meta-llama-pirated-books/675063/?utm_ source=pocket_collection_story
Russell, D. (1995) The journalistic autobiography: How reporters view themselves and order their experience, Studies in Popular Culture, Vol.18, No. 1 pp 83-99
Satariano, A. and Isaac, M. (2021) The silent partner cleaning up Facebook for $500 million a year, New York Times, 31 August
Scola v. Facebook, Inc. (2021) 18-civ-05135, Superior Court of California Count of San Mateo, 14 July
Skains, R.L. (2018) Creative practice as research: Discourse on methodology, Media Practice and Education, Vol. 19, No. 1 pp 82-97
Solon, O. (2017) Underpaid and overburdened: The life of a Facebook moderator, Guardian, 25 May. Available online at https://www.theguardian.com/news/2017/ may/25/facebook-moderator-underpaid-overburdened-extreme-content
Taylor, R. (1842) Translator’s notes to an article on Babbage’s Analytical engiro, Scientific Memoirs, Vol. 3 pp 691-731
‘Thumbs down’ (2021) 60 Minutes Australia, 17 October. Available online at https:// www.youtube.com/watch?v=xKjlMW0lQNY
Tronto, J.C. (1993) Moral boundaries: A political argument for an ethic of care, New York, Routledge
Turing, A.M. (1950) Computing machinery and intelligence, Mind, Vol. 59, No. 236 pp 433-460. Available online at https://doi.org/10.1093/mind/LIX.236.433
Acknowledgements
This paper is the culmination of work made possible by people and organisations spanning at least three continents over the past six years beginning with the fieldnotes taken in 2018 in Austin, Texas. That became the thesis finished in 2021 in Las Vegas, Nevada, and the work done on this paper began when I submitted the first draft of the abstract to the Ethical Space CFP for this issue in 2022 and finished it in 2024 in Jarrell, Texas, less than an hour away from Austin and where my formal education began over 24 years ago in Georgetown, Texas, at Southwestern University (SU). Thank you, SU, for introducing me to the academic disciplines, artistic crafts, creative/scholarly works and research methods that set the foundation for how I research, create and live: specifically, professors and fellow students in the English, Philosophy and Religion, Sociology and Anthropology, and Theatre departments. Thank you, Texas State University, and the professors in the English Department’s Technical Communication programme for showing me the vast connections and dimensions language and communication modalities have to rhetoric as a discipline – both in theory and practice – going back thousands of years. I want to thank the professors in the Creative Writing International programme and in the English department and writing center at University of Nevada, Las Vegas. I am eternally grateful to UNLV, along with the Graduate College, College of Liberal Arts, Graduate and Professional Student Association, and Black Mountain Institute giving my work and me an opportunity when no other programme would, as well as the generous financial support via scholarships, assistantships and grants during the COVID-19 pandemic. Speaking of which, I must also thank the English Department’s Composition Programme, Office of Equal Employment and Title IX, Disability Resource Center, Student Counselling and Psychological Services and Writing Center for affirming my first efforts to formally advocate for accommodations that made the better-late-than-never completion of my work and programme psychologically possible. Thank you to the editors of Hinterland magazine housed at University of East Anglia in Norwich for publishing my first piece of work on this topic. And thank you to the directors and communication team at Foxglove in London for publishing and helping me edit my last published piece on this topic and for your tireless advocacy efforts on behalf of content moderators and tech workers worldwide and helping to hold big tech companies like Meta and OpenAI accountable to the public. And a big thank you to Dr Ben Stubbs at the University of South Australia and the editors at Ethical Space for working with me on editing and revisions of this paper for the past two years despite distance and delays. Finally, I want to thank the content moderators I served with in Wave 6 at Meta in ‘18 and the writers, scholars, poets, artists and teachers I served with in the Class of ’21 at UNLV’s Creative Writing International MFA programme – thank you for treating me like a human being when I wasn’t always (or maybe ever) able to properly reciprocate. And my Mom, Cherisse, thank you for being my first English teacher. And my wife and 5th grade English teacher, Alison Darr, thank you for loving me and providing the space in our home and your heart for me to be mostly human and do the work that I love in peace – I love you (and the two cats, Tau and Mellie, too).
Funding
No funding was received for the research or the publication.
Note on the contributor
Other than the portions quoted in the Dialogue and Conclusion, this paper was written and researched exclusively by the author, Spencer Darr, and the role of OpenAI’s ChatGPT-4 was limited to dialogue collaborator, editor and research assistant. He has a Master of Fine Arts in Creative Writing (Literary Non-fiction) from the University of Nevada, Las Vegas, a Master of Arts in Technical Communication from Texas State University, and a Bachelor of Arts in English from Southwestern University in Georgetown, Texas. His writing has appeared in Atlas Obscura and Hinterland magazine with media appearances in 60 Minutes Australia, Law360 and The New York Times.