Paper short abstract:

This paper proposes a new form of digital anthropological research called "latent space ethnography", which explores the latents spaces of AI systems to better understand the emerging entanglements between datasets, AI algorithms and the social, political, and cultural dynamics embedded in them.

Paper long abstract:

One area of artificial intelligence that has gathered interest in recent years are generative models, which have been widely used to produce new forms of representations including deepfakes, AI art, and other kinds of synthetic media. Models in this broad family include generative adversarial networks (GANs) and diffusion models that generate images based on text prompts. Generative models work by identifying “latent spaces”, representing the implicit patterns in the large datasets these models are trained on, and based on which new representations can be generated.

This paper proposes latent space ethnography as a way to reconfigure anthropological approaches to AI. The anthropological study of AI has produced significant new insights into the workings of ‘black box’ AI systems, including critical analyses of predictive policing, facial recognition and algorithmic bias. These critical approaches, however, have generally approached these new powerful systems from the outside - that is, as objects of anthropological inquiry. This paper describes the opportunities that such new generative models offer when considered not only as objects of study but also as tools affording new research methods. Through latent space ethnography, we produce and engage with generative models, probing their latent space in various ways to understand the entanglements of data, AI and social/political dynamics.

In our case studies, we ask these models to envision key historical events to explore the way in which these events have been visually reconstructed in the “distributed representations” of new generative AI models.

Paper short abstract:

Hashtag brings with it a multiplicity of meanings with respect to socio-cultural and political factors of the 'local'. This enables anthropological research to develop a multi-sited research design that spans regions and time, aided by ML classification to study on a 'global' social media dataset.

Paper long abstract:

Unknown circumstances necessitate improvisation in the known framework. Conducting research on a digital network, this paper entails tracking the discourses and social transactions that take place within the network. Social media facilitates the transmission of individualistic experiences, and connects people from far and wide places through shared realities. Sharing experience includes passing on ‘broken’ meanings, and contextualising ‘whispers’ within our own time and space. This rate of flow of information is followed using multi-sited ethnographical framework, with an added skillset of ML analysis of the usage of hashtag #challengeaccepted.

What begins as a simple tagging or labelling of a post/message on the social media platform (using the hashtag), evolves into a larger collective discourse, with which others start engaging. Individuals who share their personal narratives become connected to the global context and become relatable to everyone else, in their own localised way. Hashtag activism gives a voice to the minority and the subordinated, where discourses are frequently polarised by the narratives of the majority, promoting plurality.

However, the larger question that the research paper addresses is how much social media is representative of the entire population. In the context of developing and under-developed nations, there are so many unheard and unregistered cases. The digital representation is farther away from the total. This bias is answered through lived experiences and anthropological perspective, knowing that data is triangulated by going a step further into qualitatively evidencing it.

Paper short abstract:

Can a machine learn participant observation? Could we soon see anthropologist-AI hybrids, or autonomous anthropological programmes, who will chart the new social landscapes of the metaverse? This paper will discuss the possible affordances of AI for anthropological fieldwork.

Paper long abstract:

Advancements in AI have the potential, through social media and beyond, to drastically alter human sociality. Already they are transforming the possibilities for surveillance – this in a literal, policing sense, but also in a Foucauldian sense. Thus, technologically advanced societies will become ever more deeply integrated with core ‘cyber’, even autonomous, systems. It makes sense that the anthropologists studying such societies will thus need to become adept in working with AI.

As facial recognition software grows in sophistication, it could pick up on micro-expressions during an interview, cross-reference mannerisms against stored footage, even suggest potential memetic origins therefor. In polities with mass surveillance such as China, or on platforms with vast amounts of livestreaming, eg. TikTok or Twitch, or personal data, eg. Instagram, Facebook, there is the potential for a new kind of anthropology armed with enormous data-sets and smart machine learning.

There is necessarily a debate here about privacy and permissions, but where working through algorithms alone, or only with metadata, this may be sidestepped. There are some obvious uses in surveillance and policing for algorithms trained with anthropological insights, say, to monitor complex multi-ethnic regions and social networks. Other potential uses involve the use, for instance, of medical and psychiatric anthropologists in developing algorithms that monitor users’ mental and physical health within a platform.

This paper articulates the potential of an algorithmic anthropology, its possible uses, and the ethical – indeed, some of them existential – questions it raises for the discipline as a whole.

Paper short abstract:

Considered as performance instead of science, AI embedded in small devices provides partial perspectives and incomplete knowledge, shifting attention from functionality to actors and actions, thus suited to ethnographic study. Conversely AI becomes a participant-observer of human cultures itself.

Paper long abstract:

My research is motivated by two drivers of change in AI research: a need for ethnographic approaches, and increasing technical development pushing AI into the edges of the network.

Considered as literature, craft or performance instead of a science makes AI particularly suited to ethnography. This shifts attention from the uses and outcomes of an AI deployment to the actors involved. Rapid technical development of AI, in parallel with shrinking hardware and faster processing speeds, means that both sensing and processing can take place at the edges of a network, not only in centralised servers. Accordingly, AI becomes an internal quality of things instead of their external supervisor: lightweight devices each provide partial perspectives and incomplete knowledge, together having the potential for emergent collective intelligence which incorporates both human and nonhuman actors.

More broadly, this means that AI is increasingly and invisibly embedded in natural and cultural systems. This opens up opportunities to study empirically the theory that intelligence can be seen a feature of an ecosystem. What happens when a nonhuman participant is an informant and a collaborator in the construction of knowledge?

When AI is embedded deeply and invisibly in human cultures, it could enable new perspectives on such cultures. There is growing discussion about post-humanist approaches in HCI and anthropology, but this is so far almost entirely theoretical, drawing from science fiction; a pervasive AI deployment, developed and analysed using an ethnographic approach, raises the possibility of evaluating such an approach empirically.