Extended Abstract

Artificial intelligence has long drawn inspiration from the collective behavior of ants, leading to algorithms such as Ant Colony Optimization (ACO) and decentralized decision-making frameworks in multi-agent systems. However, despite ants serving as points of reference within computational models, their complex ecologies are often oversimplified primarily to enhance computational efficiency, optimization, and human-centered utility, frequently bypassing critical ethical considerations. This paper challenges these tendencies by centering the ethical and epistemological implications of AI when applied to non-human collectives. We explore two intersecting questions: first, (a) how the historical modeling of ant behavior has informed AI through reductionist and efficiency-driven paradigms; and second, (b) how AI might be reimagined through design practice as a means of ethically monitoring ant colonies in experimental habitats.

We address question (a) in a field study titled “Seed. Potato. Pixel.”, in which compared the normative, pre-established models found in literature against real-world foraging behaviors of ants. “Seed. Potato. Pixel.” is a series of three experiments—each experiment investigates a different heuristic approach to collective foraging behavior: “Seed” explores the navigational strategies of ants; “Potato” translates ant foraging to human scale and sensory experiences; “Pixel” translates ant foraging algorithmically with digital agents. These three layers are brought together in an artistic installation to create a visual interplay between humans, non-humans, and algorithms.

Using a combination of real-time computer vision tracking, algorithmic simulation, and formicarium design, “S.P.P.” functions as an experimental interface for multispecies comparison. Our aim is not to study ants to optimize human systems, but to cultivate shared living arrangements in which ants, humans, and AI agents are entangled in ongoing negotiations of care, attention, and responsibility.

To address (b), we created a set of ontological / normative questions generated through our design and discussion sessions:

• How do we determine consent in non-human actors?
• How could we define Armeisenwürde—ant dignity—and how might it be respected?
• How do we bridge sensory and cognitive gaps in multispecies comparison/collaboration?
• How should AI systems be designed when inspired by, or intervening in, complex living systems?

We conducted a stakeholder analysis to examine the implications for ants, humans, and algorithms within our research: Ants benefit from a safe, monitored environment but face constant observation, enclosure, and reliance on human caretakers. With more agents than humans or the ACO, ants must forage for seeds to survive and are likely unaware of the experiment. Humans, aware they are playing a human-designed game, engage in relational learning and shift from data extractors to caregivers and co-designers; they can choose to ignore or break rules, as survival is not at stake. Algorithms (ACO), often seen as neutral, actively shape perceptions and actions, but unlike ants or humans, expend no energy, cannot escape the game’s rules, and have no concept of survival.

From this, 6 ethical principles arose from the design and development of “S.P.P.”, each linked to specific design and research decisions:

Mutual Care and Asymmetric Exchange

The system proposes a form of cultural symbiosis through either mutualism or commensalism: ants are provided with hydration, food, and stable living conditions; humans receive data, insight, and the opportunity to engage with non-human intelligence. However, this exchange is inherently asymmetrical. Ethical care in our project is not an abstraction—it is enacted through daily monitoring, documented maintenance schedules, and group-shared labor. This challenges the assumption that ethics can be designed in advance; instead, it unfolds through infrastructural and affective commitment.

Surveillance and Ant Privacy

Computer vision enables non-invasive, 24/7 monitoring of ant behavior. While minimizing physical disturbance, this raises questions about observation without retreat: Is constant visibility a form of care or control? Can ants be said to have a right to privacy, or opacity?

Responsibility and Long-Term Commitment

Ethical experimentation with living beings entails commitment beyond the temporal scope of research. We only proceeded with adopting ants after securing long-term rehoming agreements among collaborators. This extends ethical responsibility beyond the project’s duration and disrupts academic tendencies to treat animals as temporary research assets. The act of adoption, care rotation schedules, and anticipatory planning all shaped the infrastructure of the project and informed its epistemic boundaries.

Reframing the Lab: Wild, Pet, or Urban Co-habitant?

Our ants occupy a space between pet and lab animal, between wild and domestic. This ambiguous status opens up critical reflections on the spatial ethics of laboratory design. Our formicarium was built with escape-proof features to ensure the ants remained non-invasive to the local ecosystem, while also accommodating the colony’s needs for space and complexity. By designing a habitat that exists within a shared studio space, we reframed the lab as a domestic site, an ethical move away from extractive scientific models.

Modeling and the Ethics of Simulation

We used AI not only to track but to simulate ant behavior. This raises questions: do these models represent ant cognition, distort it, or produce something entirely new? Our simulations were instrumental in monitoring health patterns and foraging rhythms, and in predicting behavioral anomalies (e.g., lack of movement indicating dehydration). However, these models also flattened the richness of ant life into actionable metrics. We argue that simulation is not a mirror of nature, but a co-creative act with epistemological and ethical consequences.

Our project argues that ethics must be grounded in situated, relational practices, not abstract guidelines. The presence of ants in our workspace—visible during meetings, discussed in care logs, and incorporated into design decisions—continually shifted the scope and focus of the research. Ethics was not a backdrop but a shaping force. In this way, AI acted bidirectionally: it extracted data, yes, but it also returned insight, structure, and support back into the care system.

We conclude by suggesting that AI can support ethical multispecies relations if it is used not to automate away responsibility, but to distribute and extend it. For a small team managing a colony of over a thousand individuals, AI made otherwise unsustainable ethical labor possible. It enabled responsive care at scale, but only when embedded within a system designed for commitment, presence, and mutual attunement.

Keywords

Artificial Intelligence, Ant Colony Optimization, Animal Ethics, Interspecies Collaboration, Collective Behavior

Read the symposium publication here: AISB-IACAP Convention 2025 PDF.

Relationships

• Related to grANT canyon
• Related to Seed.Potato.Pixel