Using Anthropological Materials to Intervene in AI Debates
Dr Teodor Zidaru
Department of Social Anthropology
What motivated you to use AI in this module, and what goals or challenges were you aiming to address?
I considered but decided against using LLMs when teaching “Sorcery and Conspiracy: The Anthropology of Alternate Realities”. I considered it mainly out of concerns for student employability at a time when generative AI appears set to transform the nature of work and the kinds of skills students will be expected to have in their professional lives. Some degree of genAI literacy and an ability to reflect on whether and how to use AI technologies will obviously benefit students. This is especially the case in anthropology, a discipline whose students (and educators) have historically struggled to articulate clear connections between learning experiences and non-academic professional practice.
However, within the context of this particular module, using AI in the classroom risked undermining scope for in-depth engagement with course materials and instead turning AI itself into the focal point of discussion. Students had joined the module expecting a focus not on AI, but rather on magic, conspiracy theories, and associated tensions with normative understandings of ‘reality’. Sure, we could have – for example – asked an LLM to produce a conspiracy theory or a misinformation campaign. But even if we had swiftly overcome increasingly tighter guardrails against LLM misuse with a sly jailbreak or workaround, chances are the subsequent discussion would have primarily revolved around LLMs, their misuse, or their compromised training data. Conspiracy theories would have taken a back seat.
To avoid this outcome as much as possible, I discussed rather than used AI in the classroom. Moreover, the broad learning objective became understanding how course materials which ostensibly have very little to do with AI can serve as a basis for critical, independent, and original interventions in public and academic debates on AI and their social consequences.
How did you design or adapt the assessment, and how did you prepare students for using AI appropriately?
I did not redesign or adapt any assessment and students taking this module did not use AI in the classroom. I did prepare students for discussing AI drawing on course materials. To this end, I designed two weeks of teaching that touched on AI and digital technologies more broadly: one on data science and deep-learning in particular as a form of divinatory sorcery alongside horoscopy or oneiromancy; and the other on the two-way relationship between digital technologies and occult phenomena (e.g. how 4chan enables speculation about hidden conspiracies; how suspicions of hidden betrayals inspire designs for smartphone applications). These lectures helped prepare students for seminar discussions on AI technologies as always embedded in cultural systems of symbols and meanings (rather than mere tools to be used); and as novel but not unprecedented techniques for predicting and establishing orientations towards the future. Building on these insights, we closed the final seminar with a discussion on what current AI capabilities portend for the future of higher education.
What challenges did you encounter, and how did you address them?
Using LLMs as part of a seminar activity on conspiracy theories was impractical due to guardrails against misuse which activate when ‘conspiracy’ or cognate terms and phrases are used. My own inexperience with LLMs and the relative dearth of relevant examples of AI use in pedagogy were further barriers. Most examples of pedagogical AI use assume that any teaching with AI must be primarily about AI. Above all, however, it is unclear what – if anything – LLMs offer to students and teachers of anthropology. LLMs have neither an understanding of the world, nor the capacity to discern the qualitative, subjective, cultural and social dimensions of lived experience, a core anthropological skill. This skill is best developed in close independent and dialogical engagement with anthropological and ethnographic texts. Using LLMs to summarise, analyse, or compare such texts raises a range of risks, from copyright infringement to cognitive dependence and normative thought.
What benefits did you see for students and for your own teaching practice?
Students enjoyed the opportunity to engage with contemporary public techno-optimist discourses on AI and critiques thereof from an unlikely standpoint: accounts of beliefs in witchcraft, magical practices, and suspicions of conspiracy. This standpoint cleared fresh ground for intervening in these discourses in a way that appreciates the continuities between AI-based knowledge production and other ways of revealing the unknown, without falling back on neither AI hype about its ‘magical’ properties nor denunciations of the irrationality of deep-learning models (which produce knowledge based on apparently spurious correlations rather than causal frameworks). Thus prefaced, our end of module seminar discussion on the future of AI-assisted higher education created scope for healthy dissensus on which future scenarios are likeliest.
How did you evaluate the usefulness of this assessment to ensure that it reflected the desirable learning outcomes?
The rich dissensus we finished our final seminar with was, for me, a sign of the independent and original thinking that I sought to foster by relating anthropological studies of sorcery and conspiracy theories to the sociotechnical change spurred by AI and data science. We discussed, in equivocal terms, the extent to which AI may either level the playing field and make higher education more accessible to neurodivergent students or otherwise bind and ensorcell all students into cognitive dependence; the extent to which AI may misinform, mislead, or distract students from the nuances of the texts they read; or whether it is possible to divorce ‘ethical’ AI use from structural issues such as environmental costs and the precarity of pedagogical labour.
What would you do differently next time, and what advice would you give to colleagues?
At the moment, I do not think AI can be used to teach most anthropology courses, but I recognise that pedagogical AI use can be valuable in other disciplines and perhaps even in anthropology courses on technology. I would advise university management teams to not lay undue pressure on their colleagues to use AI in teaching regardless of its discipline-specific relevance. Fellow anthropologists may wish to seize current debates on AI use in higher education and broader debates on associated changes in labour markets and knowledge economies as an opportunity to highlight the strengths of anthropological thought and its irreproducibility.