Research
The question underneath all of it is what happens to painful experiences after they are over, and why people facing comparable adversity recover so differently. I have asked it about health threats, about victims and offenders, and about failure. Most recently I have asked it in AI-supported care, which is one application of the question rather than where it ends.
Research interests
Individual differences in response to adversity
Why comparable threat, failure, or harm produces such different outcomes, and what predicts which way a given person goes. This runs from my undergraduate work on dispositional threat orientation through my dissertation on behavioral responses to failure, where I modeled goal orientation, psychological capital, and intrinsic motivation as predictors of what people do after something goes wrong.
Coping, support, and recovery
What actually helps, for whom, and under what conditions. I am less interested in whether an intervention works on average than in who it fails.
Human-AI relational psychology
Whether interaction with an AI activates the relational and evaluative systems people use with each other, what that implies for how these systems should be assessed, and where the risks sit. This is my current empirical context and the source of most of my active work.
Current projects
Purpose and Execution of Memory: Foundational Differences Between Human and AI Memory Architectures, and Implications
Human memory forgets, distorts, consolidates, and reconsolidates, and those are features rather than defects. Artificial memory does none of it, because it was built for what was technologically achievable rather than for what is psychologically healthy for the person on the other end. This paper compares the two architectures directly and asks what happens to people over time when a system that never forgets and never revises becomes the place they take the things that matter most. The implications run in three directions at once: toward clinicians who will see the results, toward the people designing these systems, and toward a field that has not yet decided this is its problem.
AI and the Activation of Human Relational Psychology
Relating to an AI is not a human relationship and it is not a parasocial attachment either. Parasocial bonds are one-directional, aimed at someone who does not know you exist. An AI answers. It is responsive, personalized, and continuously available, and it supplies none of the reality-grounding that another person brings to a relationship simply by having their own perspective. We argue this is a third mode of relating with its own dynamics, and we trace how specific design features function as supernormal stimuli for evolved relational systems, drawing on attachment theory, social relationship theory, neurobiology, and self-cognition. The practical payoff is that the design choices determining whether that activation helps or harms are identifiable, and most of them are currently being made without anyone treating them as psychological decisions.
Applied behavioral research at FrayaTech
GetBackup is a live AI-supported mental wellness platform, and it is also the research environment for most of my current work. I designed the pre/post instruments for the alpha and beta cohorts, run the consent and data privacy frameworks, and code the longitudinal interaction data for behavioral patterns, engagement drivers, and clinical risk indicators. Risk detection with clinical thresholds and response protocols is built into the system architecture rather than bolted on after.
Methods
Quantitative
Structural equation modeling (MPlus), SPSS, Smallest Space Analysis, survey instrument design, longitudinal data management, SQL and PostgreSQL.
Qualitative and mixed
Systematic coding of large interaction datasets, taxonomy and framework development, mixed-methods design.
Research administration
IRB proposal development with human subjects data, informed consent frameworks, data privacy protocols, APA manuscript preparation.