Nathan Caruana, Aitor Miguel Blanco, Emily S. Cross • PsyArXiv
Finding: Physical robot co-presence confers no measurable advantage over screen-based delivery for adult language learning, with learners expecting robots to fulfill multiple social roles (coach, tutor, companion) and interact adaptively.
Why it matters: Directly relevant to HCI/UX design of educational robots, learning sciences, and the broader question of when AI/robot embodiment matters for engagement and learning outcomes.
Method: Preregistered between-subjects experiment (N=40) with a Pepper humanoid robot, comparing physical vs. screen-based conditions using objective engagement metrics, Godspeed questionnaire, gaze coding, and reflexive thematic analysis of interviews.
Jargon: *Embodied copresence* = physical presence of a robot in the same space as the learner; *Godspeed questionnaire* = validated scale measuring robot perception (anthropomorphism, animacy, likeability, etc.); *Self-in-Other scale* = measure of social affiliation/identification with the robot.
⭐ Pre-registered ⚠️ Very small sample (n=20) | Single study (not replicated) | Limited statistical detail in abstract
Finding: Preparatory sketching functions as a cognitive visual thinking system—not just mechanical recording—enabling compositional restructuring, self-assessment, and creative problem-solving in fine art.
Why it matters: Directly relevant to creativity, visual thinking, and sketching research, and supports pedagogical arguments for integrating sketching into design/arts curricula.
Method: Descriptive-analytical case study of preparatory drawings by six master artists (Michelangelo, da Vinci, Degas, etc.), examining fluency, originality, and problem sensitivity in visual ideation.
Jargon: *Chiaroscuro* = use of light/shadow contrast in art; *compositional restructuring* = reorganizing visual elements for balance and expressiveness.
⚠️ No control group mentioned | Single study (not replicated) | Limited statistical detail in abstract
Finding: Both digital and non-digital educational games improved math and geography scores in Grade 1 students, but with no statistically significant difference between conditions (d = 0.12–0.18); engagement was high in both, with offline games showing slightly stronger participation.
Why it matters: Relevant to games/gamification and learning sciences, particularly for teaching about game-based learning design, though the small sample (N=50) and weak effect sizes limit generalizability.
Method: Quasi-experimental pretest–posttest with two intact classes over 6 weeks; independent t-tests and correlation analyses used to compare conditions.
Jargon: *Self-determination theory* — motivation framework centered on autonomy, competence, and relatedness as basic psychological needs; *intact classes* — pre-existing classroom groups rather than randomly assigned participants.
⚠️ Small sample (n=50) | Self-reported data only | No control group mentioned | Single study (not replicated)
S. Haryanti, Muhammad Japar, E. Salsabila • Semantic Scholar
Finding: A contextual problem-posing learning model integrated with "hypercontent" technology improves vocational mathematics learning outcomes with medium effect sizes and high teacher acceptance.
Why it matters: Tangentially relevant to learning sciences and educational design, but narrowly focused on vocational math curriculum rather than HCI, creativity, behavior change, or team dynamics.
Method: Quasi-experimental with N-gain scores, paired t-tests, and effect size analysis using Dick and Carey instructional design framework within educational design research (EDR).
Jargon: *Problem posing* = students generate their own math problems rather than just solving given ones; *Hypercontent* = interconnected digital content with multiple media types and navigation paths; *N-gain* = normalized learning gain measuring improvement from pre- to post-test relative to possible improvement.
⚠️ Self-reported data only | No control group mentioned | Single study (not replicated)
Finding: People show consistent individual differences in willingness to monetize personal experiences, with women refusing to price tasks more than men, though no gender differences emerged in the amounts requested when they did engage.
Why it matters: Tangentially relevant to behavioral economics and decision-making around value, but the focus on gender differences in monetary valuation of identity/memory is quite removed from HCI, learning, creativity, or behavior change design.
Method: Online survey with 529 participants rating minimum compensation for 10 hypothetical scenarios (ranging from temporary restrictions to permanent identity changes), with option to refuse pricing entirely.
⚠️ No control group mentioned | Single study (not replicated)
Finding: This tutorial covers computational methods for recommending optimal team compositions using graph neural networks, focusing on automated expert selection and team refinement in online collaborative networks.
Why it matters: Tangentially relevant to team dynamics and collective intelligence research, but the approach is purely algorithmic/computational rather than behavioral or human-centered.
Jargon: *Graph neural networks (GNNs)*: ML models that learn from graph-structured data (e.g., social networks); *subgraph optimization*: finding the best subset of nodes (experts) within a larger network graph.
H. Haryanto, Larventy Larventy, Johny Budiman • Semantic Scholar
Finding: Seven behavioral biases (overconfidence, anchoring, mental accounting, loss aversion, regret aversion, herding, representativeness) affect crypto investment decisions, with financial literacy moderating only overconfidence and loss aversion effects.
Why it matters: Tangentially relevant to behavioral science and decision-making, but focused narrowly on financial markets rather than HCI, learning, or behavior change design contexts the professor covers.
Method: PLS-SEM with 520 Indonesian crypto investors via convenience-sampled online questionnaires.
Finding: Developing an interactive video-based e-book for middle school physical education using the ADDIE instructional design model improved student motivation and learning outcomes.
Why it matters: Marginal relevance to learning sciences and gamification/interactive media, but the study is narrowly applied (PE for Grade VII), methodologically weak (R&D/ADDIE with expert validation and questionnaires), and lacks behavioral or cognitive depth.
Method: ADDIE model (Analysis, Design, Development, Implementation, Evaluation) — a standard instructional design framework.
Ni Made Citariani, K. Agustini, I. K. Suartama et al. (6 authors) • Semantic Scholar
Finding: An AR framework combining brain-based learning principles with cultural values (Tri Kaya Parisudha) improved preschoolers' prosocial behavior, critical thinking, and communication in a quasi-experimental study.
Why it matters: Marginally relevant to games/serious games and learning sciences, but the focus on early childhood character education in an Indonesian cultural context is quite narrow and distant from the professor's core interests.
⚠️ No control group mentioned | Single study (not replicated) | Limited statistical detail in abstract
Finding: The Deep Dialogue-Critical Thinking (DDCT) instructional model significantly outperformed direct instruction for 10th-grade math students learning trigonometric ratios (Cohen's d = 1.18).
Why it matters: Marginally relevant to learning sciences, but the narrow focus on high school trigonometry instruction with a small sample (n=42) limits generalizability to the professor's core interests.
Method: Quasi-experimental pretest-posttest design with N-Gain scoring and Cohen's d effect size.