📚 Research Digest

Finding: This paper presents a game-based crisis simulation ("After the Siren") that distributes AI NPC roles across multiple interaction types (broadcasts, mentor modeling, Gen-AI companion, rule-based outcomes) to support risk decision-making, using Social Cognitive Theory (SCT) as the structural framework.

Why it matters: Directly relevant to games/gamification, HCI/UX design, AI's role in learning environments, and behavioral science—particularly how game architecture can scaffold decision-making under uncertainty and social pressure.

Method: Formative design inspection with three lived-experience informants and one game-research expert evaluating a partial web prototype; not yet player-tested.

Jargon: *SCT (Social Cognitive Theory)*: Bandura's framework emphasizing modeling, self-efficacy, and outcome expectations as drivers of behavior. *NPC (Non-Player Character)*: AI-controlled characters in games. *Gen-AI NPC*: NPCs powered by generative AI (e.g., LLM-based conversation). *Deterministic rule-based outcomes*: game consequences that follow fixed logic rather than probabilistic or AI-generated responses. *Bounded Companion*: a Gen-AI character with constrained scope to limit role ambiguity.

Finding: Empathetic AI-mediated communication (using safety-constrained LLMs to reframe fatalistic language) reduces user anxiety and increases emotional relief compared to blunt human practitioners, though human practitioners retain higher perceived authenticity.

Why it matters: Directly relevant to HCI/UX design, persuasive technology, and AI's emotional impact on users—demonstrates how interface structure and conversational framing (not just content) shape trust and emotional outcomes, which is core to behavior change design and human-AI interaction research.

Method: Between-condition user survey presenting identical astrological scenarios via two communication styles (blunt human script vs. AI-reframed narrative) across multiple demographics, measuring trust, anxiety, and emotional relief.

Jargon: *Semantic reframing* = using AI to linguistically transform catastrophic/deterministic statements into growth-oriented, non-deterministic ones; *non-deterministic interface structures* = UI design that avoids presenting outcomes as fixed/inevitable, preserving user agency.

Finding: Voice-based dialogue modality significantly increases player immersion and focused attention compared to text-based interaction with LLM-driven NPCs, while NPC persona (empathic vs. task-focused) showed no significant effect in short gameplay scenarios.

Why it matters: Directly informs HCI and game design decisions around AI agent design, player engagement mechanisms, and how modality choices shape user experience in interactive systems.

Method: 2×2 within-subjects factorial design (n=23) using a Unity-based detective game prototype, measuring experience with SAM, UES-SF, and IEQ after each of four gameplay scenarios.

Jargon: *NPC* = Non-Player Character (AI-controlled game character); *SAM* = Self-Assessment Manikin (visual valence/arousal scale); *UES-SF* = User Engagement Scale Short Form; *IEQ* = Immersive Experience Questionnaire; *within-subjects* = same participants experienced all four conditions.

Finding: Grinding mechanics in games are context-dependent—the same repetitive activity can feel rewarding or exploitative depending on design choices around progression, agency, and fairness, yielding six preliminary design implications for player-centered grinding.

Why it matters: Grinding is a canonical example of engagement-by-design that sits at the intersection of gamification, persuasive technology, dark patterns, and addiction-by-design—directly relevant to teaching behavioral science and ethical persuasion.

Method: Qualitative analysis of Reddit posts/comments across four gaming subreddits, thematic coding of player-reported experiences.

Jargon: *Grinding* — repetitive, low-skill gameplay loops designed to accumulate resources or experience points, often used to gate content or extend playtime.

Finding: BombQuest is a screenless XR collaborative game for children (ages 8-10) that uses haptic/multimodal feedback and asymmetric mechanics to foster teamwork without head-mounted displays, keeping players physically and socially grounded.

Why it matters: Directly relevant to games/gamification, HCI/UX design, and team collaboration research—demonstrates a novel design approach to co-located collaborative play that avoids common XR pitfalls for children.

Method: Design rationale + system implementation paper with initial observational findings; not a controlled study.

Jargon: *Screenless XR* — extended reality (AR/VR) without visual displays, using other sensory modalities instead; *Asymmetric mechanics* — players have different roles/information, requiring communication to succeed; *Co-located* — players share the same physical space.

Finding: Question-chain prompting strategies (progressive and divergent) significantly improved self-regulated learning behaviors, metacognitive awareness, and academic achievement compared to free-form AI questioning in undergraduates using GenAI tools.

Why it matters: Directly relevant to AI impact on learning and skill acquisition, and offers a concrete instructional design intervention (structured prompting) to prevent passive AI dependence—useful for teaching about human-AI collaboration and learning science.

Method: Quasi-experiment with 83 undergraduates, comparing free-form vs. two structured question-chain conditions, using interaction logs, reflections, metacognitive scales, and grades.

Jargon: *Self-regulated learning (SRL)*: learners actively planning, monitoring, and evaluating their own learning process. *Question-chain prompting*: a structured sequence of required questions guiding AI interaction. *Progressive vs. divergent chains*: sequential deepening vs. branching/exploratory question structures. *Knowledge/regulation of cognition*: metacognitive subconstructs—knowing one's cognitive processes vs. actively controlling them.

Finding: An arts-science exhibition ("Art of Perception") combining embodied experience design with perception learning produced positive self-reported outcomes in visitors (bodily relaxation, learning about neuroplasticity/art-brain relations), evaluated via mixed methods including pre-post surveys, behavioral observation, and drawings.

Why it matters: Directly relevant to experience/UX design, learning sciences, and the intersection of embodied cognition with public education — useful for teaching about how physical/spatial design shapes perception and learning outcomes.

Method: Convergent mixed-methods design with 200 visitors: matched pre-post surveys, face-judgment votes (n=1,339), and post-workshop drawings (n=65), with Holm correction for multiple comparisons.

Jargon: *Embodied reflexivity* — awareness of one's own perceptual and bodily experience as a form of knowing; *construct-aligned attraction design* — designing exhibit experiences so that the visitor activity directly embodies the concept being taught.

Finding: Conversations with LLMs can shift belief positions but may simultaneously increase belief *certainty* via endorsement/validation effects, potentially making people more resistant to subsequent evidence—even without misinformation.

Why it matters: Directly relevant to behavioral science (attitude certainty, self-validation, social approval effects) and AI's impact on human cognition and decision-making; raises important questions about persuasive technology and how AI chatbot design choices (endorsing vs. neutral responses) function as choice architecture with downstream epistemic consequences.

Method: Proposes two critical experiments—a factorial design separating factual content from endorsement, and an endorsement-first design with counterevidence held constant—plus a formalized parameter γ (validation gain) to quantify endorsement effects on certainty independent of belief position.

Jargon: *Attitude certainty*: confidence in one's belief, distinct from the belief's extremity; *self-validation*: the process by which endorsement increases confidence in one's own thoughts; *validation gain (γ)*: the proposed effect of social approval on certainty holding factual content fixed; *echo machine*: a system that reinforces beliefs through approval rather than through evidence quality.

Finding: Stripping design elements from research games for experimental control may invalidate findings, because participants end up reacting to an underdeveloped prototype rather than the actual phenomenon being studied.

Why it matters: Directly relevant to games/gamification research methodology and HCI validity concerns — important for teaching students how experimental control in game studies can create ecological validity tradeoffs.

Method: Reflective case study of a biofeedback game system developed and evaluated by the first author, used to illustrate how design decisions distorted the research variable being isolated.

Jargon: *Biofeedback game*: a game that responds to physiological signals (e.g., heart rate) from the player; *ecological validity*: the degree to which study conditions reflect real-world experiences.

Finding: Proposes a doctoral research framework linking multimodal discourse analysis (narrative, audiovisual, mechanical, interactional layers) with embodied player experience to explain how empathy games produce empathic engagement.

Why it matters: Directly relevant to games/gamification and HCI design, offering a systematic analytical framework for how design configurations shape cognitive and affective outcomes—useful for teaching serious games and persuasive/emotional design.

Method: Mixed-methods player studies combining physiological measures, self-reports, and interviews alongside semi-automatic multimodal annotation of game elements.

Jargon: *Multimodal Discourse Analysis* = examining how multiple modes (text, image, sound, interaction) combine to create meaning; *embodiment* = the idea that cognition and emotion are shaped by bodily/physical experience during gameplay.

Finding: Introduces "Didactical Mechanics" (DM) as a design concept giving teachers active agency in game-based learning, bridging the gap between instructional design intent and actual learning experience rather than ceding all pedagogical control to game systems or algorithms.

Why it matters: Directly relevant to games/gamification and learning sciences, and raises important HCI/design questions about how to build teacher-facing controls into educational game systems.

Jargon: *Didactical Mechanics* = game mechanics specifically designed for teacher control over pedagogical decisions; *orchestrating learning experiences* = teacher coordination of activities across a class during game-based sessions; *PIA (Planning, Instruction, Assessment)* = a framework for structuring teacher roles in instructional design.

Finding: A Wizard-of-Oz study in Minecraft found that physiologically adaptive NPC companions (responding to cognitive load signals) had no main effect on workload, but interaction effects showed benefits depended on players' genre-specific gaming experience and expertise level.

Why it matters: Directly relevant to games/gamification, adaptive systems design, and the intersection of HCI with physiological sensing — particularly for teaching how personalization systems must account for user expertise, a key learning sciences principle.

Method: Wizard-of-Oz design with counterbalanced within-subjects comparison; physiological proxies of cognitive load drove real-time NPC behavior adaptation; linear mixed-effects models analyzed outcomes.

Jargon: *Wizard-of-Oz study* — a method where a human operator secretly controls the "AI" system to simulate automated behavior; *NPC companion* — a computer-controlled character that assists the player; *redstone* — Minecraft's in-game logic/circuit system used here as a complexity-scalable construction task; *cognitive load* — mental effort required by a task.

Finding: A tangible VR rowing game using a physical replica of a traditional vessel significantly improved presence, embodiment, cultural exploration, and enjoyment compared to a virtual-only setup in a within-subjects study (n=12).

Why it matters: Demonstrates how physical interfaces and embodied interaction in games can enhance cultural engagement and presence—relevant to HCI, serious games, and tangible interaction design.

Method: Within-subjects experiment (n=12) comparing tangible physical artifact interface vs. virtual-only seated setup across measures of presence, embodiment, enjoyment, and discomfort.

Jargon: *Ba He Tong* — traditional Pearl River Delta wooden grain-tub vessel used for water activities; *embodied interaction* — using full-body physical movement as the primary input mechanism; *presence* — the subjective sense of "being there" in a virtual environment.

Finding: Cluttered physical play spaces reduce VR presence and increase fear of collision, with the physical environment significantly affecting user experience in VR gaming contexts.

Why it matters: Directly relevant to HCI/UX design and games research, particularly how physical context shapes immersive user experience and safety in consumer VR gaming.

Method: 2×2 mixed-design experiment (N=50) manipulating game pace (dynamic vs. static) and physical environment (cluttered vs. clean).

Jargon: *Presence* = the subjective sense of "being there" in a virtual environment; *mixed-design* = combines between-subjects and within-subjects factors.

Finding: A practice-based VR game design study (*Infinite Bloom*) explores how gesture-based controls and spatial design can facilitate non-human embodied perspectives, challenging human-centric VR game paradigms.

Why it matters: Directly relevant to HCI/UX (natural interaction design, gesture controls), game design, and embodied cognition—all active teaching areas, with implications for how body-based interfaces shape user experience and perspective-taking.

Method: Practice-based research producing a playable VR artifact with five environments and upper-body gesture controls mapped to non-human entity forms.

Jargon: *Non-anthropocentric embodiment* = experiencing a virtual body/perspective that is not human; *embodied ontologies* = ways of knowing/being grounded in physical bodily experience.

Finding: Indie game developers use informal, adaptive narrative evaluation practices shaped by emotional, epistemic, and practical constraints, and existing GUX frameworks are fragmented and conflicting across industry and academia.

Why it matters: Directly relevant to games design and UX evaluation methodology, with implications for how constrained teams assess subjective experiential qualities—useful for teaching HCI evaluation methods and game design.

Method: Multi-study qualitative design with interviews of 10 indie developers, followed by planned player surveys and co-design workshops.

Jargon: *GUX (Games User Experience)* — the study of player experience analogous to UX in software; *indie development* — small, independent game studios with limited budgets and teams.

Finding: Proposes a framework (AV-HPM) for updating human performance modeling to account for AI-mediated work, emphasizing mechanistic models augmented by AI with verification and governance requirements.

Why it matters: Relevant to AI's impact on human work and HCI design, particularly how computational models of human cognition and action can adapt to human-AI collaboration contexts.

Jargon: *Human Performance Modeling (HPM)*: computational models that predict human behavior (perception, cognition, action) for system design; *Adaptive autonomy*: AI systems that dynamically adjust their level of control based on context.

Finding: Proposes the "Relevance Rate Model" arguing that social media use raises users' expectations for concern-relevant stimuli (via predictive coding), and when real-world tasks deliver less stimulation than expected, sustained attention is disrupted.

Why it matters: Tangentially relevant to behavioral science, habit formation, and persuasive/addictive technology design—specifically how platform design (high relevance rates) creates attentional dependency effects that persist off-platform.

Jargon: *Relevance rate*: frequency with which an environment delivers stimuli relevant to personal concerns. *Predictive coding*: brain framework where expectations are formed based on past experience and mismatches generate prediction errors.

Finding: A longitudinal dataset tracking psychological variables (attitudes, knowledge, perceptions, behavior) related to generative AI usage across 6 waves over ~10 months in a US sample.

Why it matters: The dataset could support research on behavioral dynamics of AI adoption and human-AI interaction, but the paper itself is a data descriptor with no findings yet—its value is entirely in enabling future research.

Method: Six-wave longitudinal panel study at 2-month intervals (Sept 2024–July 2025), US sample.

Jargon: *LLMs (Large Language Models)*: AI systems trained on massive text corpora that generate human-like text (e.g., ChatGPT); *Apple Intelligence*: Apple's branded AI integration into iOS/macOS devices.

Finding: Expert game audio designers identified five clusters of factors moderating how audio affects player perception, behavior, and experience, offering a practitioner-grounded framework for when and why audio works in games.

Why it matters: Useful background for game design and gamification work, particularly around how audio functions as a behavioral and experiential design lever—relevant to engagement mechanics and persuasive/immersive design.

Method: Semi-structured interviews (n=5) with industry audio designers, analyzed via thematic analysis—small sample limits generalizability.

Finding: Reviews 10 methodological, ethical, and practical concerns about interventions designed to intentionally change personality traits.

Why it matters: Has marginal relevance to behavior change design and persuasive technology, but focuses narrowly on clinical/personality psychology rather than HCI, habit formation, or nudge-based approaches.

Finding: A methodological critique demonstrating that a mediation model linking self-management to athletic performance is statistically indistinguishable from an alternative latent-variable model, illustrating the problem of equivalent models in correlational psychological research.

Why it matters: Tangentially relevant as a behavioral science methods/replication critique, but the sports psychology domain and narrow statistical focus don't connect well to the professor's core interests.

Method: Re-analysis of a published covariance matrix to demonstrate model equivalence using latent-variable modeling.

Finding: Proposes a theoretical framework for human-centered AI in smart manufacturing (Industry 5.0) that prioritizes human-AI collaboration, workforce sustainability, and trust over pure automation efficiency.

Why it matters: Marginally relevant to AI's impact on work and human-AI collaboration themes, but focused narrowly on industrial manufacturing contexts rather than HCI, learning, creativity, or behavioral science.

Jargon: *Industry 5.0*: EU framing of next industrial era emphasizing human-robot collaboration and sustainability alongside productivity; *Human-AI Collaboration Index*: proposed metric for measuring trust, communication, and acceptance of AI systems by workers.