Finding: Systematic review of 43 empirical studies on AI interventions in science education, finding AI tools (generative AI, adaptive learning, immersive environments) support critical thinking, collaboration, and sustainability competencies.
Why it matters: Tangentially relevant to learning sciences and AI impact on education, but the focus on SDG 4/sustainability framing and science education specifically is far from the professor's core interests in HCI, creativity, teams, behavioral science, or games.
Method: Systematic review following PRISMA 2020 guidelines across 43 studies (2020–2026).
B. Divjak, B. Svetec, Damir Horvat • Semantic Scholar
Finding: This scoping review defines "learning design quality" (LDQ) and demonstrates how learning analytics can evaluate course designs before and after implementation, finding educators should align designs with contemporary pedagogies.
Why it matters: Peripheral relevance to learning sciences, but focused narrowly on instructional/course design quality assurance in higher education rather than skill acquisition, HCI, behavior change, or creativity.
Jargon: *Learning design (LD)*: structured plan of teaching/learning activities; *Design analytics*: data-driven analysis of the design artifact itself (pre-implementation); *A priori/a posteriori evaluation*: assessment before vs. during/after course delivery.
Pablo Martínez-López, Francisco Garre-Frutos, Miguel A. Vadillo et al. (4 authors) • PsyArXiv
Finding: Systematic review of 14 experimental paradigms for studying habit formation in humans, finding most fail to demonstrate valid habit formation, with the forced-response task being the most reliable measure (reliability .74–.83).
Why it matters: Directly relevant to behavior change design and habit formation research — this review exposes the methodological weakness underlying much of the science that persuasive technology and behavioral intervention design relies on.
Method: Systematic review plus reanalysis of publicly available datasets to assess psychometric reliability of habit measures across paradigms.
Jargon: *Dual-system framework* — theory that behavior is controlled either by deliberate goal-directed processes or automatic habit systems; *outcome-insensitive* — behavior that persists even when the outcome/reward is devalued, the hallmark test of a habit; *construct validity* — whether a measure actually captures the theoretical construct it claims to measure; *forced-response task* — a paradigm where response speed pressure is used to elicit automatic vs. deliberate responses.
Finding: VR-based historical narrative immersion (transportation) correlates positively with emotional responses and story-consistent political beliefs but not factual knowledge, using the Anne Frank House VR as a case study.
Why it matters: Directly relevant to games/serious games, persuasive technology, and behavior change design—specifically how immersive media shapes attitudes and engagement without necessarily improving knowledge.
Method: Correlational study (N=106) using the transportation-imagery model to measure narrative immersion, emotional response, knowledge, and political belief alignment.
Jargon: *Transportation* = narrative immersion, the degree to which a user is mentally "transported" into a story world; *story-consistent beliefs* = attitudes aligned with the implicit message of the narrative (e.g., pro-refugee sentiment after an Anne Frank VR experience).
Chenxi Li, Xiaoxue Hu, Xue Wang et al. (4 authors) • PsyArXiv
Finding: Using Google Trends data linked to 1.5M player-day records across 86 countries, higher national attention to AI predicted increased voluntary helping behavior between online game players, with a quasi-experimental boost around OpenAI DevDay.
Why it matters: Directly relevant to AI's impact on human social behavior and collaboration, and uses an online game as a behavioral measurement platform—relevant to both the AI-society and games/gamification interests.
Method: Novel quasi-experimental design linking country-level Google Trends AI attention to in-game prosocial behavior, with natural experiment around a major AI event (OpenAI DevDay).
Jargon: *Player-day records*: unit of analysis combining individual players and daily observations; *quasi-experimental*: exploits a natural event rather than random assignment to estimate causal effects.
⚠️ No control group mentioned | Single study (not replicated) | Limited statistical detail in abstract
Kenneth J. Kokroko, Meredith E. Hovis, A. Zuniga-Teran et al. (13 authors) • Semantic Scholar
Finding: A pilot gamification-based engagement tool ("Dream Garden Game") successfully supported community participation in urban greening policymaking by having participants design hypothetical gardens, yielding qualitative data on plant preferences and cultural drivers.
Why it matters: Directly relevant to gamification and serious games design, particularly how playful design mechanics can drive civic engagement and behavior change in real-world policy contexts.
Method: Participatory pilot study using an interactive card/design game followed by group discussion and critical reflection on the methodology itself.
Jargon: *Ecosystem services* — benefits humans receive from natural environments (e.g., shade, air quality, stormwater management); *playful design* — incorporating game-like or whimsical elements into non-game contexts to increase engagement.
⚠️ No control group mentioned | Single study (not replicated)
Finding: Pre-service teachers who played tabletop RPGs alongside youth shifted their conceptions of teaching, learning to support differentiated instruction, productive conflict, and more equitable, social-justice-oriented game-based learning.
Why it matters: Touches on games/serious games, learning sciences, and skill acquisition in teacher education, with some relevance to how game-based contexts shape collaborative dynamics and goal-setting with youth.
Method: Social design experiment using pre/post interviews, teaching artifacts, reflections, and field notes to trace conceptual shifts in three pre-service teachers.
Jargon: *PST* = pre-service teacher; *multiliteracies pedagogies* = frameworks for teaching that recognize multiple modes of communication and cultural contexts; *agentic shifts* = changes in how individuals exercise intentional, self-directed action within their roles; *TTRPG* = tabletop roleplaying game.
⚠️ No control group mentioned | Single study (not replicated) | Limited statistical detail in abstract
Eileen Richter, Lukas Wesenberg, Felix Krieglstein et al. (7 authors) • PsyArXiv
Finding: Pre-practice instructional support (exploratory or structured "blackboards") in a digital math platform reduces error rates compared to practice-only conditions, based on large-scale log data from 9,293 learners.
Why it matters: Touches on learning sciences and skill acquisition in digital environments, with some relevance to how platform design shapes learning outcomes — tangentially relevant to HCI and behavior change through interface design.
Method: Within-subject design using large-scale log data (74,801 problems), comparing four conditions of instructional support timing and type.
Finding: Post-error behavioral adjustment in chess depends on whether the error outcome is perceptible (material loss visible on board) vs. invisible (engine-only detectable blunder) — visible errors produce speeding up, invisible ones produce slowing down, across 6.7M moves.
Why it matters: Touches on behavioral science and decision-making (error monitoring, metacognition, self-regulation) with implications for feedback design in learning and behavior change systems.
Method: Large-scale naturalistic behavioral data (1.06M event observations, 222K games) with within-person analysis across skill tiers — strong ecological validity.
Jargon: *Post-error slowing* — the tendency to respond more slowly after making a mistake; *blunder* — a chess move that significantly worsens the position per engine evaluation; *material* — chess pieces with point value (losing them is perceptible).
Finding: Post-error behavioral adjustment (slowing vs. speeding) depends on whether the error outcome is perceptible — visible material loss in chess causes faster subsequent moves, while invisible errors cause slowing, suggesting two distinct error-monitoring mechanisms.
Why it matters: Relevant to behavioral science and decision-making, particularly error monitoring, metacognition, and how feedback visibility shapes behavior change — concepts applicable to learning sciences and persuasive technology design.
Method: Large-scale naturalistic behavioral analysis of 6.7 million chess moves across 222,099 games, using engine evaluation to distinguish perceptible vs. imperceptible errors — a clever separation of error occurrence from error noticing.
Jargon: *Post-error slowing* — the well-documented tendency to slow down after making a mistake; *blunder* — a move that significantly worsens a player's position per engine evaluation; *material loss* — losing pieces, which is visually obvious on the board.
Finding: Anthropomorphism ratings of robot faces did not predict trustworthiness perceptions, and individual differences in tendency to anthropomorphize did not mediate this relationship—all three hypotheses were null results.
Why it matters: Tangentially relevant to HCI and human-robot interaction design, particularly for understanding how robot physical design influences user acceptance, but the null findings and narrow focus (face trustworthiness ratings, n=43) limit practical design implications.
Method: Small-sample (n=43) correlational/mediation study using Godspeed and IDAQ questionnaires with 30 robot image stimuli.
⚠️ Small sample (n=43) | Self-reported data only | No control group mentioned | Single study (not replicated)
Ni Made Citariani, K. Agustini, I. K. Suartama et al. (6 authors) • Semantic Scholar
Finding: An AR-based educational framework combining brain-based learning principles with Balinese cultural values (Tri Kaya Parisudha) improves prosocial behavior, critical thinking, and communication in early childhood learners.
Why it matters: Touches on learning sciences, gamified/interactive educational technology, and behavior change design in early childhood contexts, but the cultural-specific framing and early childhood focus are peripheral to the professor's core interests.
Method: ADDIE instructional design model with quasi-experimental evaluation and expert validity review.
Jargon: *Brain-based learning (BBL)*: instructional approach aligned with neuroscientific principles of how the brain learns; *Tri Kaya Parisudha*: Balinese Hindu ethical framework emphasizing purity of thought, word, and action.
⚠️ No control group mentioned | Single study (not replicated) | Limited statistical detail in abstract
Ahsan Ali, Waseem Bahadur, Lu Liu et al. (4 authors) • Semantic Scholar
Finding: AI-HRM digitization affects employee performance through two opposing pathways—job-related cynicism (negative) and innovative job behavior (positive)—with task discretion moderating which pathway dominates.
Why it matters: Touches on AI impact on work and behavioral responses to technology, but is primarily an organizational behavior/HR management study with limited connection to HCI, behavior change design, or the other core interests.
Method: Three-wave time-lagged multisource survey (N=543 employee-supervisor dyads in China), allowing causal ordering while controlling for common method bias.
Finding: Public trust in AI clinical psychology decision-making is shaped by transparency, bias concerns, privacy, regulation, and human oversight, with people preferring AI as assistant rather than autonomous decision-maker.
Why it matters: Tangentially relevant to AI's impact on human collaboration and trust, but focused narrowly on healthcare/clinical psychology rather than HCI, behavioral science, or the professor's core teaching areas.
⚠️ No control group mentioned | Single study (not replicated)
Finding: The paper argues that "durable skills" (communication, critical thinking, teamwork) are actually domain knowledge in use, and that competency modeling frameworks obscure this by collapsing knowledge into behavioral labels.
Why it matters: Tangentially relevant to learning sciences and skill acquisition, but the focus is workforce taxonomy and job analysis methodology rather than HCI, behavioral science, creativity, or team dynamics.
Justin Wahlers, William Hart, Oh-Ryeong Ha et al. (5 authors) • PsyArXiv
Finding: The Invested and Demanded Effort Model (IDEM) distinguishes between effort people *choose* to exert (invested) versus effort *required by circumstances* (demanded), showing these have opposite relationships with motivation and self-control in nicotine abstinence contexts.
Why it matters: The invested/demanded effort distinction has marginal relevance to behavior change design and persuasive technology — particularly for designing quit-support tools or understanding user effort perception — but the paper is narrowly focused on smoking cessation rather than HCI or behavioral design.
Jargon: *Invested effort* = effortful control attributed to one's own volition; *Demanded effort* = effort attributed to external self-control demands (e.g., cravings); *IDEM* = theoretical framework distinguishing these two effort types.
⚠️ No control group mentioned | Single study (not replicated) | Convenience sample (e.g., MTurk, students) | Limited statistical detail in abstract
Finding: Proposes "Accountable Design" as a framework for integrating legal values (privacy, security, trust, safety) into human-centered design processes for AI and IoT technologies.
Why it matters: Tangentially relevant to HCI and design ethics, but focuses primarily on legal accountability rather than the core topics of creativity, behavior change, teams, goal-setting, or games.
Teeramate Jirawutthipan, Thada Jantakoon, S. Trisupakitti • Semantic Scholar
Finding: Bibliometric analysis of 331 papers on VR in nursing education (2018-2024) reveals rapid growth (37.32% CAGR) but a pre-consolidated field lacking strong central themes, with outcome measurement focused on simulation rather than clinical transfer.
Why it matters: Tangentially relevant to HCI (VR/immersive learning cluster identified) and learning sciences, but the primary focus is nursing education curriculum—not design, behavior change, or team dynamics.
Method: Bibliometric science mapping using Biblioshiny and VOSviewer for keyword co-occurrence networks and strategic thematic diagrams.
⚠️ No control group mentioned | Single study (not replicated) | Limited statistical detail in abstract
G. Tiwow, I. K. Arsana, H. Dolonseda • Semantic Scholar
Finding: TPACK integration mediates the relationship between digital platform utilization and learning effectiveness among high school teachers, explaining 60.7% of variance in learning effectiveness.
Why it matters: Marginally relevant to learning sciences and technology integration in education, but focused on a narrow regional survey of Indonesian teachers rather than HCI, creativity, behavioral science, or the professor's core interests.
Jargon: TPACK (Technological Pedagogical Content Knowledge): framework describing teachers' ability to integrate technology with pedagogy and subject content; PLS-SEM: structural equation modeling approach for testing mediation pathways.
Finding: A theoretical health economics model showing that price-based interventions (e.g., taxes on unhealthy goods) can paradoxically reduce healthy behavior due to reinforcement linkages between health stock and consumption habits.
Why it matters: Has marginal relevance to behavioral science and nudge/choice architecture research, particularly regarding unintended consequences of price-based behavior change interventions.
Jargon: *Rational addiction model*: economic framework where agents optimally choose addictive behaviors with full foresight; *adjustment costs*: friction costs for changing behavior levels; *comparative statics*: analysis of how equilibrium outcomes change with parameter shifts.