Research
I study AI and digital technologies as constitutive parts of evolving social systems, tracing how interactions among people, technologies, and information produce community- and institutional-level outcomes.
01
How do people and intelligent machines shape one another in social systems?
My research on human–AI interaction is part of the ERC-funded HUMANET project. Across observational and experimental studies, I examine how automated agents become social participants, acquire recognizable roles, and reshape trust, behavior, and discourse.
ERC-funded HUMANET01
Preprint · IC2S2 2026
This project maps community-identified bots across Reddit to trace their roles and long-term evolution. It shows that while community-developed bots declined, their functions persisted as automation became concentrated in platform infrastructure, revealing a shift from decentralized participation toward centralized governance.
Read the preprint02
Ongoing · Related work presented at ICA 2024
I examine human–human, human–bot, and bot–bot encounters to understand how bots operate as community characters and moderators. The project asks how interactions with bots influence the trajectory, targets, and topics of subsequent human discourse, including incivility.
03
Ongoing experiment · CS2Nordics 2026
Using an online multiplayer bluffing game, this project tests whether people deceive, distrust, and reason differently when they believe they are interacting with machines. By holding behavior constant while varying perceived identity, the experiment separates responses to machine identity from responses to machine behavior.
02
How does narrative structure shape what people remember, adopt, and pass through social networks?
This program studies narratives as structured systems of people, events, states, and relations. I examine how those structures interact with prior beliefs and social contexts to influence the interpretation, memory, transformation, and diffusion of information.
Developed through NSF-funded Online Dynamics of Misinformation
01
Under review · IC2S2 2026
In a networked Story Loom experiment, participants assembled stories from information elements and passed them through social ties. The results show that causal connectedness predicts adoption more strongly than sequence, with effects becoming more pronounced as narratives develop; prior beliefs shape ideological content while leaving much of the underlying relational structure intact.
02
Under review · IC2S2 2026
This work represents news narratives as cumulative graphs of actors, actions, states, and causal relations. Combining human annotation with large language models, I study how narrative agency, responsibility, and turning points differ across news environments and relate to sharing and engagement.
03
Manuscript in preparation · ICA 2023
Across vaccine, SIDS, and folktale materials, this project investigates how internal narrative coherence and compatibility with background beliefs shape recall. It shows how memory can preserve causal organization while merging episodes in culturally patterned ways.
03
How do individual behavior, social relationships, and community contexts produce collective patterns online?
This work treats online communities as dynamic systems in which individual tendencies, network positions, group norms, and platform contexts continually interact. It examines how disruptive behavior emerges, spreads, and changes over time rather than assuming it is simply a stable trait of particular users.
Foundational research program
01
Entropy · 2025
Using computational mechanics, this study models trolling as a set of recurrent behavioral transitions. It reveals how disruptive behavior can spread through interaction and how the same person may move among different behavioral states over time.
Read the paper02
Computers in Human Behavior · 2021
A social network analysis of YouTube communities shows how relationships and network positions shape responses to trolling. Users embedded in dense communities and occupying central positions are more likely to respond, underscoring the role of group structure and norms.
Read the paper03
Frontiers in Political Science · 2021
Drawing on eleven years of Reddit data, this study compares incivility across political, non-political, and mixed communities. It finds that incivility varies by context and responds to external events and platform policies, while its overall proportion remains relatively stable over time.
Read the paperAcross the work
Across these projects, I combine large-scale behavioral data, social network analysis, statistical modeling, computational text analysis, machine learning and large language models, agent-based modeling, and controlled and networked experiments. My future research brings human–AI interaction and computational narrative analysis together to examine how machine identity and behavior shape trust, cooperation, narrative production, and shared meaning in mixed human–AI networks.