Summer 2025 Graduate Interdisciplinary Research Accelerator Fund (GIRAF) Recipients

This project aims to develop a constructive theory about what organisms can do when faced with new problem situations they are not presently equipped to deal with, and which they do not understand. Specifically, it aims to defend an hypothesis similar to one advanced by the founding Gestalt Psychologists that intelligent organisms, when faced with a problem situation, have some capacity to directly experience the structural requirements of the situation; and that this experienced structure can be sufficient to guide and drive the discovery or invention of a successful solution even if the organism has no history of successfully solving similar problems before. The Gestalt psychologists called this direct experience of structural requirements insight. Unlike the Gestalt theorists, this

dissertation aims at a theory of direct insight. That is to say, in a similar vein to James Gibson’s ecological theory of direct perception, these experienced features of problem situations are real structural features of the organism’s life situation, and the process of developing insight is a process of the organism progressively reintegrating itself into its life situation until it can effectively coordinate with it, as opposed to, for instance, structuring a mental representation of the world in its head.

Bowen has recently submitted a journal article on this subject, describing the issue and its attendant research opportunities; this GIRAF grant supports the next two papers in a series (the three papers will, with modification, comprise the core of his dissertation). The first article transitions from description to explanation and asks: What is the causal basis of insight and its development in problem situations? This paper will reconstruct the theory given by the Gestalt psychologists, and the productive research strategy they followed. However, their theory involved positing two worlds–a “phenomenal” problem situation represented in the head which is directly experienced and a “real” situation which is only mediately and indirectly experienced. He argues that this gives rise to a kind of Dualism, the Two World Problem, which is problematic in all of the familiar ways. This leads to the second journal article, which will show how James and Eleanor Gibson adapted the Gestalt framework in a manner explicitly designed to avoid the Two-World Problem. In the process, they undercut a core premise of Koffka’s argument that the meaningful structure we experience has to be built up out of proximal stimuli which lack this meaningful structure. In place of this scheme, James Gibson invented ecological optics, the concept of an affordance, and a reconceptualization of the perceptual process as one of information pickup. By means of these concepts, the Gibsons developed a strategy for relocating the perceived meaning and value in the organism-environment system rather than inside the organism, and they found a way to render perception direct. But, Anderson argues, both James and Eleanor Gibson’s approach and contemporary ecological psychology are currently incapable of accounting for crucial features of insight.

Communication, such as making promises about future actions, are fundamental to forming trust in both formal and informal human relationships. Behavioural experiments have consistently demonstrated that humans often put aside self-interest in order to keep their past non-binding promises. Predominant  explanatory mechanism for such human trust relies on the negative reinforcement of psychological guilt or guilt-aversion resulting from untrustworthy behaviour. However, generative AI systems lack neurobiological responses such as guilt, raising critical questions about the mechanism of trust formation in human-AI interactions. As AI systems become integral to our social and economic environments, it is imperative to investigate how trust is established between humans and AI, and how it differs from human-human interactions.

The goal of the project is to examine the role of communication in trust formation across three relational contexts: human-human, human-AI, and AI-AI interactions, and across three simulated situational contexts of economic, social, and robot interactions. The study employs a multifactorial approach with the relational and situations contexts as the main factors. The anticipated outcomes of this project is three fold:

1. experimental findings that highlight the differences in human and generative AI communication for the purposes of trust formation

2. peer-reviewed journal submission detailing the findings on the nature of these differences

3. Fostering international collaboration with European partners, setting the foundation for future EU Horizon grant applications.

The study uses a mixed-method approach with an adaptation of the repeated trust game as the core experimental paradigm. Participants in a repeated trust game communicate in natural language in order to establish and maintain trust in a repeated play of a strategic game where there is an incentive for one participant to be untrustworthy due to self-interest. By analyzing the nature of the communication and their relation to the strategic choices made by participants, Bowen and Baruah are establishing the role of communication in trust formation.

Experimental design: Since contextual factors influence trust formation, they are adapting the repeated trust game to three contextual settings:

1. Semi-formal economic partnership, which simulates trust dynamic resulting form verbal contracts

2. Informal social interaction, which simulates everyday casual interactions

3. Robot interaction, which simulates trust formation between a human and a simulated household robot. The experiment is also being conducted in two arms:

Control Arm: This arm establishes the baseline human trust data for each of the contexts using the oTree framework. Participants for the informal and human-AI contexts will be recruited through the Prolific crowdsourcing platform, while existing literature will provide data for the economic partnership context.

AI arm: Each context is replicated with large language model (LLM)-based generative AI agents assuming participant roles.

Analysis: The project is being undertaken with a mixed-method sequential analysis, beginning with qualitative analysis followed by the quantitative analysis. Bowen and Baruah use thematic analysis for qualitative data of the natural language communication, and the resultant themes act as independent variables in the quantitative analysis of the strategy choices.

The fundamental goal of artificial intelligence development is to perfect/optimize performance in whatever domain it is operating in (Manyika et al., 2023). E.g., AI in vehicles is meant to make driving error-free, efficient, and comfortable; AI in healthcare is meant to improve treatment outcomes/diagnostic accuracy; AI on content-streaming platforms is meant to improve the quality of recommendations for each user. Stark and Chevalier characterize the optimization mindset as one of the core factors driving AI innovation such that engineers and researchers begin by asking what is flawed about the domain they plan to improve through AI and then design an AI to combat those flaws.

This year, there have been several high-profile cases of AI for friendship and community. Robot friends have long been the domain of science fiction, and it is not surprising that with the introduction of effective generative AI, people have begun to think of how AI can finally be applied to human relationships. Stark and Chevalier see the premise of AI friendship development to be similar to the premise for all AI development that there is something in human-to-human relationships that needs optimizing. They note that the reasoning behind certain high-profile friendship bots like Avi Schiffmann’s friend [sic] is that the developers have found traditional relationships unsatisfying. More people are reporting that modern life disconnects them from a community of friends, a phenomenon dubbed the “loneliness epidemic” by journalists. Others are finding that human relationships can be exhausting, cruel, painful, and unfulfilling on their own leading to a slew of influencers promoting a “soft life” free from romance, children, and any relationship demands that might disturb the peace. For these people, friendship AI improves on the parts of human relationships that lead them to feel lonely and hurt. An AI friend is always available yet never demanding, always supportive yet never suffocating. At most, the cost of an AI friend is a simple monetary payment.

The main aim of this project is thus to critically evaluate the promises of AI for relationships and respond to the above claims that AI can one day offer an optimized relationship experience. Stark and Chevalier are defining what a successful relationship between two humans might look like, and how AI might be able to replicate or fail to replicate those traits. The argument that AI can fix or avoid the flaws of human relationships is a strong one, and to respond they are looking to the premise that human error is something that needs improving. Instead, they are exploring the innate value of flawed human relationships. They are reviewing relevant literature across philosophy, sociology, computer science, and bioethics.

While there is increasing importance ascribed to interdisciplinary scholarship and its practice, the discussion to date has strongly focused on 1) how “traditional” University structures (e.g., Departments, Faculties) might change to better support interdisciplinary scholarship and 2) intra-science interdisciplinary scholarship, to the exclusion of other fields and ways of knowing. The philosophy of science interrogates the value of interdisciplinarity particularly (but not exclusively) through the study of Scientific Pluralism (https://plato.stanford.edu/entries/scientific-pluralism). Furthermore, philosophy of science is itself a case study of interdisciplinary research, and, importantly, is a case study that transcends science.

This project is developing the basis for a programme of research on the conversation between the philosophy of science and interdisciplinarity, where Lizotte defines the latter as inclusive of the full breadth of scholarly disciplines, to investigate how scholars “do interdisciplinarity” when outside traditional academic structures.

This project is conducting a background study to kickstart this new research project. The main activity undertaken by Demirkaya is a two-part review and synthesis of 1) current best practices and strategies for supporting and carrying out interdisciplinary research in non-traditional organizational structures such as interdisciplinary institutes and institutes for advanced study, and 2) work within the philosophy of science scholarship, starting from scientific pluralism, that might speak to these best practices.

This project aims to challenge the anthropocentric framing of epistemic agency in contemporary (and historical) epistemology and philosophy of mind by exploring how both humans and artificial intelligence (AI) can be appreciated and understood as machines that think. Anthropomorphic AI interfaces, particularly those of consumer-facing large language models, reinforce an outdated, human-centric conception of cognition by encouraging users to attribute beliefs, desires, and intentions to AI. These interfaces capitalize on folk conceptions of mind and intelligence, shaping how users perceive AI as something that either approximates or falls short of human cognition, rather than as a distinct kind of cognitive system. By rejecting both anthropocentrism (the privileging of human cognition as the standard for intelligence and mindedness) and anthropomorphism (the [mis]attribution of human-like qualities to non-human entities), this project argues for a more precise conceptualization of epistemic agency. The project proposes adopting an interactionist approach, seeing epistemic agency as emerging from interactions between a cognitive system and its environment, including other cognitive systems. This shift moves us beyond viewing human-AI interaction as either human-tool dynamics or imitation of human cognition, instead recognizing it as an engagement between distinct but comparable epistemic agents each with unique affordances and limitations.

This reconceptualization is essential for several reasons:

First, it clarifies epistemic agency as an interactive, emergent phenomenon. Rather than assuming humans possess intrinsic epistemic superiority, we should analyze how epistemic agency arises through interaction between different cognitive systems -biological and artificial. Understanding knowledge practices in this way will help design AI systems that complement human cognition rather than merely reproduce it.

Second, it prevents epistemic distortions caused by anthropomorphic design. When AI is framed as a pseudo-human, users are prone to over-trusting, misunderstanding, or misattributing its capacities. A more precise, non-anthropomorphic and non-anthropocentric framing enables lay users to interact with AI in ways that recognize its distinct modes of reasoning and information processing.

Third, it promotes a more rigorous approach to human-AI epistemic collaboration. If we acknowledge that epistemic agency is not exclusive to humans but emerges in various forms, we can better assess AI’s role in scientific inquiry, knowledge generation, and reasoning processes. Rather than asking whether AI thinks like us, we should explore how different epistemic agents – human and artificial- can interact to achieve knowledge goals more effectively.

Finally, it reframes the ethics of AI in knowledge production. If AI is understood as an epistemic agent rather than a passive tool, then questions about responsibility, accountability, and epistemic authority must be reconsidered – are AI systems moral agents or is responsibility and accountability imbedded in the environment in which the system acts? This has profound implications for AI governance, transparency, and the role of AI in shaping public knowledge.

To advance this shift in perspective, this project is examining: (1) How epistemic agency emerges from human-AI interactions, rather than residing solely in either humans or machines; (2) The ways in which anthropocentric biases distort our understanding of AI’s epistemic role; (3) Design principles for AI interfaces that move beyond anthropomorphism, emphasizing epistemic complementarity rather than imitation; and (4) The role of AI as an epistemic agent in scientific inquiry, and whether it should be considered a collaborator rather than merely a tool.

This project, which will be part of Kucuk’s dissertation, examines how network theory can offer a new perspective on mental disorders. Traditional approaches to psychiatric classification individuate mental disorders as discrete entities based on clusters of symptoms. In contrast, network theory conceptualizes and provides the means for studying mental disorders as dynamic systems shaped by interactions between symptoms, environmental factors, and individual experiences. Kucuk aims to explore whether this theoretical shift provides a more accurate and flexible framework for understanding mental disorders and what implications it has for clinical practice and the role of patients in conceptualizing their own conditions. By investigating these questions, Kucuk’s research aims to bridge insights from philosophy of science and psychiatry, critically assessing the epistemic and methodological foundations of a network-theoretic approach to mental health.

Kucuk leads the project, undertaking philosophical analysis of network theory in psychiatry and drawing on literature on scientific explanation, the nature and modeling of mental disorders, and the increasing focus on patient-centered perspectives in psychiatry. She is developing a conceptual framework that explores how traditional symptom-based models differ from relational and process-based understandings of mental disorders. Additionally, she is analyzing how this shift affects psychiatric diagnosis, treatment strategies, and patient agency. By integrating both theoretical and applied perspectives, her research is contributing to ongoing debates in philosophy of science and mental health, offering a more nuanced understanding of psychiatric disorders and their implications for clinical practice.

Chatbots and virtual agents such as OpenAI’s ChatGPT have come to dominate both technical and societal conversations around the impacts of generative artificial intelligence (AI) systems. These technologies and their developers are disrupting various aspects of everyday life for Ontarians. This project is documenting, analyzing, critiquing, and influencing the design of contemporary interactive artificial intelligence technologies such as ChatGPT in Ontario’s education and creative sectors. A detailed examination of the design and effects of these interactive generative AI technologies through a) primary source archival, b) ethnographic interview, and c) legal/policy research in specific sectors of the Ontario economy will assist stakeholders in formulating specific institutional and policy responses to this wave of technological change.

The project supports a wider discussion of alternative design and regulatory choices for generative AI technologies to ensure these tools work for the benefit of all Ontarians. By understanding how today’s interactive AI systems are reshaping human perceptions and actions around social interaction, intimacy, and the nature of intelligence itself, the project supports recommendations and interventions for policymakers, creative practitioners, and the public at large, particularly Ontario youth who are at the forefront of both using and being impacted by these technologies.

In the first investigative phase of the project,  Takuya conducted a literature review and media monitoring survey to support further work on all three aspects of the project: a) primary source archival research b) ethnographic interviews, and c) legal/policy research. Takuya is also identifying appropriate archival collections, potential interviewees, and applicable legal materials as part of their literature review and scoping exercise.

In this project, Kat Newman is identifying and comparing the effects of environmental pollution, and climate change more broadly, upon transgender people living in the United States and Canada. There is currently no systematic research, data, or evidence on this topic. While there is an existing body of literature on gender and the environment, gender is almost always equated to the experiences of cisgender women and men. By homogenizing the experiences of LGBTQI people, the existing literature on queer ecology also fails to speak to the experiences of transgender people, which are quite different from those of other sexual and gender minorities. There is an established body of literature on environmental racism, a form of systemic racism whereby poor urban communities of colour are disproportionately burdened with health and safety hazards through policies and practices that compel them to live in proximity to infrastructures such as sewage works, landfills, power stations, and major roads and thoroughfares. Like low-income communities of color, transgender people are also subjected to poor housing conditions and polluted environments that erode their health and quality of life. Given the similar experiences between people of colour and transgender people, and the fact that approximately 40% of the transgender population in the U.S. and Canada identify as non-white, the environmental racism literature is helpful for understanding transgender experiences of climate change. However, research carried out specifically to understand transgender people’ s experiences is also necessary since trans experiences are not always identical to those of racialized communities.

Under the supervision of Dr. Baruah, Newman is reviewing and synthesizing existing peer-reviewed literature on trans experiences of climate change in fields such as geography, gender studies, sociology, and environmental studies, using online search engines such as Google Scholar, Scopus, and Thomson Reuters (formerly ISI) Web of Knowledge. The peer-reviewed scholarship on transgender experiences of climate change is currently quite limited, but there is a significant amount of working knowledge available from journalistic and practitioner sources. Newman is also collecting and analyzing relevant news articles as well as professional reports, policy reviews, position papers and survey results from trans advocacy organizations such as PFLAG, Egale, Trans Lifeline, and the Trans Wellness Initiative in Canada, and Advocates for Trans Equality, Transgender Law Centre, and the Trevor Project in the US. Collected data will be analyzed using the Codebook for Standards of Evidence for Empirical Research (SoE) (Heck and Minner 2009). The SoE and their process of application result in a careful review of the claims of individual studies and reports based on six categories: adequate documentation, internal validity, analytic precision, generalizability/external validity determination, overall fit, and warrants for claims.

Loop Quantum Gravity (LQG) characterizes dynamics as the evolution of physical variables relative to each other, without relying on an independent temporal parameter. This relational approach to time is often contrasted with classical mechanics, where time is treated as an absolute, independent variable against which all dynamical evolution occurs. However, emphasizing this contrast can obscure important continuities between classical and quantum gravitational theories. Does the absence of absolute time in General Relativity (and by extension, in LQG) imply the complete irrelevance of time as a physical concept? Or do certain structural and conceptual features of classical time persist in these modern frameworks?

This project is investigating the critical features of time as understood in classical mechanics, and is assessing whether these features persist in the conceptual framework of Loop Quantum Gravity. The ultimate goal is to understand whether a fully timeless theory is viable, or whether elements of classical temporality continue to shape our understanding of quantum gravitational dynamics.