Where does AI fit for me, in my research and scholarly practice? This is still a work in progress. I’m both grappling with the question of where AI fits in my research and scholarly practice, and I’m also trying to get a sense of what is possible with the AI tools we currently have.
Ethan Mollick says, “… because AI has a “jagged frontier” (it is not good at everything, but rather has uneven abilities) it is unlikely to do every task that a worker is responsible for. Improving the performance in a few areas need not lead to replacement, but instead it will free up time to do more valuable, satisfying, and productive work.”
How you use AI should align with your values – What is important to you? What do you value most about your research? About your scholarly practice? And, it’s important to note that what you consider valuable, satisfying, and productive may differ from what someone else considers valuable, satisfying, and productive. Centering values in the decision-making process is not trivial, especially given how those values are entwined with our identities as researchers and scholars. For some, doing a lit review or annotated bibliography entirely themselves is a core piece of how they consider themselves as a scholar, and thus, highly valued. For others, having AI assist with these tasks allows them to spend more time on other tasks that they consider more core to their identities as a scholar, and thus, more valuable. (For a terrific exploration of this idea, I recommend Butson and Spronken-Smith’s AI and its implications for research in higher education: A critical dialogue.)
On the question of where AI fits in my own practice – I’ll admit to being worried about being judged for whatever decision I make. Indeed, to paraphrase last week’s digital detox, opinions on the role of AI in research and scholarship are polarized. I think it’s important to acknowledge that in this current context of polarization, staking a claim means opening ourselves up to criticism of our values, and that is no small barrier to overcome. (This is where an institutional or disciplinary culture of transparency around AI norms and expected use would be handy.)
That said, I’ll take the plunge and share that although my exploration of generative AI in my own research and scholarly practice is still a work in progress, I have found opportunities to use generative AI in ways that feel aligned with my values as a research and scholar – so far, mostly in the areas of idea generation, writing, and editing:
- Exploring the contours of a topic (I prefer Claude for this task)
- Finding sources and summarizing sources (using tools like Research Rabbit and Ai2 OpenScholar)
- Helping to generate outlines (again, Claude)
- Providing a rough draft for an abstract, or helping to cut down words or polish an abstract that I’ve already written (and Claude again)
Regarding tools, there are an overwhelming number of generative AI tools available, and most of us don’t have the time or inclination to keep up with every possible tool. I’m trying to balance that tension by sticking with a tool that I like when I find one, and also keeping my feelers out for other options that have the potential to improve on my current setup.
Finally, I’ll share one use case that is not aligned with my values. While generative AI shows some promise for deductive text analysis of large corpuses, I’m not comfortable with using generative AI for inductive analysis. As a researcher who is drawn primarily to questions that are best explored through qualitative methods, I strongly believe in taking every opportunity possible to be “close to the data” – to have “full” knowledge of the data (an often-unproblematized, but not unproblematic, idea in and of itself, as explained by Lyn Richards in her piece Closeness to the Data: The Changing Goals of Qualitative Data Handling). I therefore have little interest in using AI tools that shortcut the methods and processes that I use to remain close to the data as the research and analysis evolves.
Beyond a misalignment with my own values, the epistemological assumptions embedded within AI systems may themselves be mis-aligned with my goals as a qualitative researcher. As Butson and Spronken-Smith (2024) write,
“While algorithms can churn through vast amounts of quantitative data, they are at a loss when nuanced understanding and contextual interpretation are required. For example, AI algorithms are ill-suited for capturing the multifaceted nature of human behaviour and social constructs (Ligo et al., 2021; Sloane & Moss, 2019). This is not just a technological shortcoming; it is a profound misalignment with the epistemological frameworks that these disciplines are built upon” (p. 566)
This reminds me of Meredith Broussard’s contention that “mathematical truth and social truth are fundamentally different systems of logic” (2023, p. 2). Technologies are not neutral, and unpacking the assumptions embedded within technological systems is part of the journey of understanding how to use generative AI in ways that align with your values and goals.
Dr. Sarah Lohnes Watulak