Abstract
Anyone who has graded a paper or reviewed a manuscript in the last year knows about generative artificial intelligence, or AI, such as ChatGPT and Microsoft Copilot. Suddenly, our work as educators, reviewers, and readers has an added coda of nail biting, head shaking, and brow furrowing. Not only do we need to decide if what we are reading makes sense and flows smoothly, if it contributes something relevant and (ideally) new, if it excites and interests us as readers, if it leaves anything out, if it follows the requirements of the assignment or submission guidelines – we also have to wonder if a person actually wrote it. Now, academic sleight-of-hand is nothing new, but the scale and skill of AI feel to most of us like nothing we have prepared for. Here we are in our collective nightmare, facing the exam we forgot to study for, and the clock is ticking.
Yes, free and readily accessible online platforms allow pretty much anyone with an Internet connection to generate text of a given length, on a set topic, in a requested style, all within seconds. While the earlier programs spit out prose that was reassuringly clunky, recent AI-generated writing is suddenly all too slick. AI can mimic student writing. AI can write like academics. AI can fool us – and we know it. Plus, let's be honest, writing is an activity that most of us have decidedly mixed experiences with, especially in the “neoliberal churn” of productivity that so many of our institutions demand (Goodkind et al., 2023). What if writing could happen in seconds – no muss, no fuss? How tempting is that?
When we took part in conference panels featuring editors for several social work journals, AI inevitably came up in the question and answer sessions. What is your policy on the use of AI in submitted manuscripts? What guidelines do you have for authors? But really, the underlying question was: are you all okay with this? We needed to start with a recognition that we are not “all” feeling the same about AI. How could we be? The editors shared our different thoughts and feelings about the role of AI in social work scholarship. Some thought it would benefit social work research. Some encouraged writers to use it for editing. Some felt uncomfortable with it and worried about ways it would perpetuate false claims or poach ideas without credit. Some expressed concerns about its effectiveness in data analysis. Then most of us expressed an earnest plan to develop more complete guidelines for authors after we had a chance to confer and consult.
The three of us then discussed our thoughts about AI with the Affilia editorial board at our annual retreat, and heard from our publisher's representative about AI policies adopted across other journals. Sage's journal policies are available for authors online (see https://us.sagepub.com/en-us/nam/artificial-intelligence-policy#:∼:text=AI%20assistance&text=AI%20tools%20that%20make%20suggestions,disclosure%20by%20authors%20or%20reviewers). Such publishing guidelines emphasize that AI cannot be a co-author and that the human authors are responsible for anything that is submitted as “their” work. In other words, we can’t blame ChatGPT for writing something misleading, misrepresenting our methods or our data, or plagiarizing others. It is also true that a hard-core anti-AI stance would be tough to pull off when today's programs for data analysis, grammar checking, and even email writing all come with embedded “AI assistance”. Research is a human endeavor, without a doubt, and yet it is also technologically enabled. AI is undeniably helpful, especially for people who are writing in a style or a language that is less familiar to them. We have long encouraged writers to seek out editorial services. We send students to campus writing centers. We use spell checkers, search engines, and predictive text options. So when does AI become a problem for us as feminists, scholars, and social workers?
Gita Mehrotra has explained that, as critical feminist social work scholars, “How we do the work is the work”; working slowly and carefully is not an optional frill but is at the core of our praxis (2023, title). Talking together as an editorial board about AI is one example of our praxis, and we will continue to develop our journal policies in deliberate and collaborative ways. We are working on a policy that will further clarify our expectations on AI use within Affilia.
For a different take, we also decided to ask generative AI – ChatGPT – about this challenge that we are facing. We started with the prompt: Write about the impact of generative AI for critical feminism and social work in the style of an editorial for an academic journal. A full editorial appeared, with nicely shaped paragraphs and appropriate section headers. It was balanced and academic in tone. It integrated social work and feminism. It was, frankly, alarming. It was also a bit long, so we requested a shorter version at 200 words. Boom. Instant summary:
The Promise and Peril of Generative AI for Critical Feminism and Social Work
Generative AI offers both opportunities and risks for critical feminism and social work. On one hand, AI can democratize knowledge, amplify feminist thought, and support social work by making research more accessible and fostering creative activism. It allows rapid data analysis, helping practitioners and scholars address structural inequalities and track policy impacts on marginalized groups.
However, AI is not neutral. It often reproduces biases embedded in its datasets, risking the reinforcement of sexist, racist, and heteronormative assumptions. This poses serious concerns in social work, where biased algorithms could lead to discriminatory practices in welfare, healthcare, and policing. AI also threatens to devalue care work, a labor-intensive field disproportionately carried by women, potentially eroding the quality of human-centered services.
To harness AI responsibly, feminist scholars and social workers must ensure ethical development, advocating for inclusive datasets, participatory design, and accountability in AI deployment. AI's potential lies in aligning its use with feminist values of justice, equity, and care. By doing so, we can prevent technology from perpetuating systemic oppression and instead use it as a tool for social liberation.
What can we see in this generated text? First of all, AI comes off looking pretty good – more “promise” than “peril” if you steer around the “serious concerns”. But more than that, the way this mini-editorial is written makes it difficult to disagree with. We probably like those verbs: democratize, amplify, support. Then there's that even, balanced appeal to “justice, equity, and care” – who is going to storm off in a huff after reading that? The content even mimics critical self-awareness and acknowledges the bias that can lurk in AI-generated work. Readers can certainly pick at the text's phrasing and question its focus, but it does more “right” than “wrong”. It guesses some fields that critical feminist social work academics would be interested in. It identifies itself as a potential threat, humbly acknowledging the importance of care work done primarily by women. Yes, by golly, let's “prevent technology from perpetuating systemic oppression and instead use it as a tool for social liberation”!
Then there is the “Author's Note.” Okay then, who is the “Author” writing about “ethical oversight”? Who is determining whether “AI supports, rather than undermines, feminist and social justice goals”? Who is “harnessing AI responsibly”? Who is “aligning its use with feminist values”? And the answer, from this text that appeared in a matter of seconds is: nobody. Nobody is authoring, nobody is supporting, and nobody is overseeing.
When the three of us first read the AI-generated editorial tidbits, one of us asked, “What are we even here for?” While this was said in jest, it might be a good way to start thinking about AI in our work at Affilia. What are we even here for? Why does it matter that we are humans when we are doing our work? Why are we taking the time that it takes to read, think, write, talk, share, listen, revise, gather, meet, confer, analyze, reflect, try, fail, read some more, write some more, make mistakes, learn some things, and unlearn some others? Our AI-generated exercise certainly shows that the keywords and stock phrases of critical feminism, of social work, and of academia are all too easy to put together in sentences, paragraphs, and arguments. But AI can’t tell us what it is here for – it doesn’t have a “why”. Why develop a pastiche of other texts, blending ideas that fall so easily into place that they won’t generate surprise, or wonder, or disagreement? Because the prompt requested it.
As people, we learn through our living bodies, over years, in the midst of our relationships with human and more-than-human others – what Donna Haraway calls “multispecies muddles” (2016, p. 31). We have to hold ourselves and each other to account if we want to embody the feminist values that AI can unfurl so rapidly across the page. We need to know the “why”. Words that we write come from many places and moments; in large datasets, it can seem like they come from nowhere at all. Only we can make sure that we stand behind our words. Positionality is about more than just listing our identities – we need to know how and why it matters that we are the people who have done what we have done (Jackson et al., 2024). When we are here as critical feminists and social workers, when we are here as writers and readers, we are here for something – so let's be clear what that is.
A glaring omission in the AI-generated text is the recognition that training AI models both consumes an incredible amount of energy and intensifies the production of carbon emissions (Budennyy et al., 2022). As critical feminist social workers and scholars, climate justice is not an optional consideration. The use of AI in the production of knowledge has climate implications, a reality that we must reckon with as artificial intelligence becomes more entangled in our work as scholars.
When you consider submitting manuscripts to Affilia, we hope that you want us to hear from you – and as readers, reviewers, and colleagues we expect that you, as a particular individual or group, are sharing work that you are proud to be doing. You are the author, sharing words you have selected, about what you have done, and what more you would like to nurture. AI can’t hold values. AI can’t be accountable. AI can’t care. You can. When you author something, you build a connection with the people who will read it. You care enough to help us all to do our own work, and we are grateful for your efforts.
As critical feminist social workers, let's write about what matters to us, in our precious, slow-moving, messy, and intertwined lives. We – as humans – are more than algorithms, more than generators of word counts on short timelines. We care about what happens next because we are already in the middle of it; we are affected and we are engaged. Otherwise, what are we even here for?