📚 Research Digest

Finding: Individual differences in achievement goal orientation (approach vs. avoidance goals) predict how the brain's reward system responds to negative feedback—people higher in approach goals show less negative ventral striatum responses to negative feedback, suggesting they perceive it as more instructive than punitive.

Why it matters: Directly relevant to goal-setting theory and behavioral science of motivation; has practical implications for how educators and designers frame negative feedback in learning environments and performance systems.

Method: fMRI during a word association learning task with performance feedback, following a competence threat manipulation—a notably rigorous neuroscientific approach to a behavioral question.

Jargon: *Ventral striatum (VS)*: brain region central to reward processing and motivation. *BOLD responses*: blood-oxygen-level-dependent signals measured by fMRI, used as a proxy for neural activity. *Approach vs. avoidance goals*: achievement goal orientation distinguishing between pursuing positive outcomes vs. avoiding negative ones.

Finding: An AI-native deck-building game uses multi-agent LLMs for card generation, NPC decision-making, and natural language interaction; a 13-person playtest showed engaging gameplay but surfaced issues with predictability, transparency, and player control.

Why it matters: Directly relevant to game design and gamification topics, and raises important HCI questions about transparency and player agency in AI-driven systems that parallel broader concerns about human-AI collaboration and control.

Method: Small-scale playtest (n=13) with a working prototype; structured JSON parsing used to constrain LLM outputs to designer-specified numerical values.

Jargon: *AI-native game* — game where core rules/content are generated dynamically by AI rather than pre-authored; *deck-building game* — genre where players construct card sets during play; *multi-agent system* — multiple AI agents operating semi-autonomously and interacting.

Finding: The intention-behavior gap in sustainable eating is reframed using active inference and ecological affordances, arguing that unsustainable behavior is a predictable response to misaligned environments rather than individual motivational failure.

Why it matters: The affordance-based framework (pragmatic value + predictive precision + habit strength) directly maps onto behavioral science interests in choice architecture, nudges, and habit formation—and the institutional/environmental intervention focus is highly relevant to behavior change design pedagogy.

Jargon: *Active inference* = a neuroscience framework where agents minimize prediction error by acting on the world; *affordances* = action possibilities that emerge from person-environment interactions; *policy prior* = habitual tendency toward a behavior encoded as a prior probability.

Finding: Self-control has a significant negative relationship with social media addiction (r = –0.505) among young adults (18–24), meaning stronger self-regulation predicts lower addictive social media use.

Why it matters: Directly relevant to behavior change design, specifically habit formation, addiction by design, and persuasive technology ethics — useful for teaching about self-regulation as a design and intervention target.

Method: Correlational quantitative study (n=107), Pearson correlation with standardized instruments.

Jargon: *Social media addiction* — compulsive, uncontrolled social media use that interferes with daily functioning; *self-control/self-regulation* — capacity to override impulses and manage behavior toward goals.

Finding: IVR instruction significantly increased immersion and agency compared to PowerPoint in undergraduate STEM courses, but did not significantly improve motivation or presence, suggesting CAMIL framework applicability depends heavily on implementation context.

Why it matters: Relevant to learning sciences and technology-enhanced instruction, with implications for how immersive technologies affect student engagement and skill acquisition.

Method: Counterbalanced crossover design with 40 undergraduates across three modules, using repeated-measures ANOVA and Wilcoxon signed-rank tests on paired IVR vs. PPT observations.

Jargon: *CAMIL* = Cognitive-Affective Model of Immersive Learning, a framework linking immersion → presence → agency → motivation in VR learning; *agency* = learner's sense of control over their learning experience; *counterbalanced crossover* = all participants experience both conditions in alternating order to control for sequence effects.

Finding: Proposes a conceptual framework positioning educational psychology as the foundation for AI-enhanced education, covering learner diversity, inclusive pedagogy, and reflective teaching in intelligent learning environments.

Why it matters: Relevant to AI's impact on learning sciences and education, but is a high-level policy-oriented framework review rather than empirical work on specific HCI, behavioral, or team dynamics questions the professor focuses on.

Method: Systematic narrative synthesis of ~620 sources (65 retained) across PsycINFO, ERIC, Scopus, and Web of Science, with cross-regional comparative analysis.

Finding: Doctoral students show high GenAI literacy but this doesn't translate to clear understanding of roles, ethics, or agency in human-AI collaborative writing, with misalignment driven more by contextual pressures than lack of awareness.

Why it matters: Relevant to AI's impact on knowledge work and learning, particularly around epistemic agency and responsible AI use in academic contexts.

Method: Explanatory sequential mixed-methods with 62 doctoral students: GenAI literacy assessment followed by semi-structured interviews and cross-case comparison to identify alignment typologies.

Jargon: *HAICW* = Human-AI Collaborative Writing; *epistemic agency* = the capacity to take ownership of one's own knowledge-building process; *GenAI literacy* = ability to critically and effectively use generative AI tools.

Finding: Dutch policymakers report that behavioral insights are inconsistently integrated into equitable public health policy, identifying five key needs including more context-sensitive knowledge and better collaboration between behavioral scientists and policymakers.

Why it matters: Tangentially relevant to behavioral science and behavior change design, but focused on public health policy translation rather than HCI, learning, teams, or persuasive technology design.

Finding: Novel assistive AI systems that split perception-action loops between human and AI create ambiguous responsibility boundaries, termed a "responsibility gap."

Why it matters: Tangentially relevant to HCI and AI-human collaboration, but focused narrowly on assistive tech ethics/responsibility for visually impaired users rather than the professor's core areas.

Finding: Staff in a democratic Sudbury school view technology as both a natural learning environment and a disruptor of "productive boredom," proposing that collective community discourse about technology constitutes a learning practice ("community heutagogy").

Why it matters: Marginally relevant to self-directed learning and digital literacy, but too focused on a niche alternative schooling philosophy to be directly useful for HCI, behavioral science, or learning sciences as typically taught.

Jargon: *Heutagogy*: self-determined learning where learners direct their own education; *Connectivism*: learning theory emphasizing knowledge distributed across networks; *Sudbury school*: democratic free school with no mandatory curriculum.

Finding: Survey of 135 construction management students identifies academic, financial, and social factors that enable or hinder sustained motivation, with practical institutional recommendations like peer mentoring and career fairs.

Why it matters: Touches on motivation and learning sciences but is narrowly focused on a single discipline (construction management) with limited theoretical or methodological contribution relevant to HCI, behavioral science, or broader learning sciences.

Finding: Integrating 5E inquiry-based learning with STAD cooperative learning significantly improved Grade 11 students' conceptual understanding of photosynthesis (d=1.88, g=0.62) in a quasi-experimental study.

Why it matters: Marginally relevant as evidence for cooperative/team learning structures and inquiry-based pedagogy, but the domain (biology content knowledge) and context (Thai secondary school) are far from the professor's core interests.

Method: One-group pretest-posttest quasi-experimental design (n=62); no control group limits causal inference.

Jargon: *5E model* = Engage, Explore, Explain, Elaborate, Evaluate instructional cycle; *STAD* = Student Teams Achievement Divisions, a structured cooperative learning method where team scores depend on individual improvement; *normalized gain (g)* = pre-to-post improvement relative to maximum possible improvement.

Finding: Proposes CARE-AI, a conceptual framework for responsible AI integration in primary education covering pedagogy, adaptive learning, ethics, and teacher empowerment — but it is explicitly unvalidated.

Why it matters: Tangentially relevant to AI's impact on learning and HCI, but the primary/K-12 focus and purely conceptual (no empirical data) nature limit utility for a professor focused on HCI, behavioral science, and adult learning contexts.