📚 Research Digest

Finding: High GenAI users offload cognitive work substantially (58% felt "AI did most of the thinking"), and greater prompt dependence correlates with reduced confidence in independent reasoning (r = −.61), with qualitative themes around lost cognitive ownership and speed-vs-depth trade-offs.

Why it matters: Directly relevant to teaching about AI's impact on cognition, decision-making autonomy, and cognitive biases—and raises design questions about how GenAI tools should be built to preserve human agency and critical thinking.

Method: Mixed-methods with 1,923 professionals completing 10 simulated executive tasks using real GenAI tools, measuring prompt behavior, override frequency, and self-reported autonomy via Likert scales.

Jargon: *Cognitive offloading* = delegating mental work to an external tool; *override behavior* = user choosing to reject or modify AI output; *executive demands* = higher-order cognitive tasks like planning, sequencing, and decision-making.

Finding: A Design and Development Research study creating a gamified Moodle course for probability education, integrating interactive videos, original HTML/JS games, and collaborative activities using the ADDIE model and constructivist frameworks.

Why it matters: Directly relevant to teaching about gamification, serious games, instructional design, and HCI in learning environments—offers a concrete, reproducible framework connecting game design, engagement, and learning science.

Method: ADDIE instructional design model applied across eight progressive modules with custom-built educational games and formative assessments embedded in an LMS.

Jargon: *ADDIE* = Analysis, Design, Development, Implementation, Evaluation—standard instructional design process model; *DDR* = Design and Development Research, methodology focused on creating and evaluating educational artifacts; *Connectivism* = learning theory emphasizing networked knowledge acquisition through digital tools.

Finding: Controllability gates the intrinsic motivational value of progress—improvement only increases enjoyment, effort, and task preference when people attribute it to their own actions rather than chance.

Why it matters: Directly relevant to goal-setting, gamification design, and motivation theory: explains why feedback systems and skill-building activities must give learners genuine agency over outcomes to sustain engagement.

Method: Two experiments (pilot + preregistered replication) manipulating progress level and controllability independently, measuring subjective enjoyment and task preference.

Jargon: *Gating hypothesis* — controllability acts as a gate that either allows or blocks progress from translating into motivational value; *cognitive control* — the mental processes regulating effortful, goal-directed behavior.

Finding: Selection bias in expertise research distorts conclusions about skill development — studying only elite performers creates misleading patterns where beneficial early predictors appear to weaken or reverse, not because the development pathway changed, but due to statistical artifacts from sampling on the outcome.

Why it matters: Directly challenges assumptions underlying expertise and skill acquisition research, with implications for how we teach about deliberate practice, talent development, and learning sciences — the "10,000 hours" framework and similar models may be built on selection-biased samples.

Method: Combines simulation modeling with longitudinal data from chess and athletics to demonstrate how collider bias and range restriction distort predictor-outcome relationships in elite samples.

Jargon: *Selection bias / collider bias* — when a sample is filtered on a variable causally downstream of your predictors and outcome, it induces spurious correlations among those predictors; *range restriction* — attenuated variance in elite samples shrinks observed effect sizes.

Finding: GPT-4o and Claude exhibit human-like cognitive biases (risk aversion, loss aversion, demand chasing) in inventory decision-making tasks, and deliberate prompt design significantly reduces these biases.

Why it matters: Directly relevant to behavioral science and decision-making interests, and raises important questions about AI as a behavioral actor—relevant for teaching AI's impact on work and human-AI collaboration in decision contexts.

Method: Controlled newsvendor problem experiments with LLMs mirroring established human-subject laboratory protocols, with systematic prompt framing variations.

Jargon: *Newsvendor problem* — a classic operations management decision task where an agent must choose inventory quantity under demand uncertainty, balancing overage and underage costs; *demand chasing* — the bias of anchoring orders too closely to the most recent demand observation rather than the statistical optimum; *bounded rationality* — decision-making that is rational within cognitive/informational limits.

Finding: Cross-national study (N=2,470) shows people increasingly attribute consciousness and emotional experience to AI, with convergence across countries toward believing AI will be conscious within five years.

Why it matters: Relevant to AI's impact on human-AI collaboration and trust dynamics, but the focus is more on psychology/governance than HCI, creativity, or behavioral science as taught in this professor's courses.

Method: Preregistered mixed-methods cross-national survey combining quantitative mind perception scales with qualitative emotional description analysis across five countries.

Jargon: *Mind perceptions*: attribution of mental states (agency = capacity to act/plan; experience = capacity to feel) to non-human entities.

Finding: Aphantasia (inability to generate mental visual imagery) shapes creativity, learning, and self-understanding in significant ways, with implications for education and creative practice.

Why it matters: Relevant to creativity and visual thinking research—aphantasia challenges assumptions about sketching, visualization, and ideation processes that underpin much HCI and design pedagogy.

Method: Thematic analysis of open-ended survey responses and email testimonies from 52 aphantasic adults (drawn from 6,436 respondents).

Jargon: *Aphantasia*: inability to form voluntary mental images; *VVIQ*: Vividness of Visual Imagery Questionnaire, standard measure of imagery vividness.

Finding: VEGAS uses eye-gaze data to evaluate and select video captions that better match individual viewers' visual attention, improving alignment between captions and human focus.

Why it matters: Marginally relevant to HCI through gaze-tracking as a user signal, but the core contribution is a computer vision/NLP evaluation metric with little connection to the professor's teaching areas.

Finding: Visual imagery vividness exists on a continuum from hyperphantasia to aphantasia, and latent class analysis reveals three distinct subtypes of aphantasia differing in sensory modality breadth, imagery intensity, and associated cognitive features.

Why it matters: Has marginal relevance to creativity and sketching/visual thinking research, since aphantasia affects designers and creative practitioners differently, but this paper is primarily a cognitive psychology/neuroscience classification study with limited HCI or pedagogical application.

Method: Latent class analysis on 3,883 online respondents using imagery vividness questionnaires and self-reported cognitive measures.

Jargon: *Aphantasia* = inability to form mental images; *Hyperphantasia* = extremely vivid mental imagery; *Latent class analysis* = statistical method for identifying hidden subgroups within a population.

Finding: Behavioral finance has matured as a discipline that extends (rather than replaces) rational-agent models by incorporating psychological biases like prospect theory, mental accounting, and herd behavior to explain market anomalies.

Why it matters: The cognitive biases and decision-making frameworks covered (heuristics, overconfidence, mental accounting) overlap with behavioral science and decision-making topics, but the application domain is financial markets, not HCI, creativity, teams, or learning.

Jargon: *Prospect theory*: people evaluate gains/losses relative to a reference point, not absolute outcomes; *mental accounting*: people categorize money into separate mental "buckets" that affect decisions irrationally.

Finding: Cooperative learning strategies significantly improve secondary school students' academic achievement in biology compared to traditional instruction, based on a quasi-experimental study in Nigerian schools.

Why it matters: Tangentially relevant to learning sciences and team/collaborative dynamics, but narrowly focused on K-12 biology pedagogy in a specific regional context with no connection to HCI, creativity, AI, or behavioral science.

Method: Quasi-experimental design with 150 students using a Biology Achievement Test (BAT) analyzed via independent t-test.

Finding: A physical Visual Geometry Kit (VGK) for teaching 3D spatial visualization (orthogonal projection, plans/elevations) to middle schoolers was developed and validated, showing high usability scores across usefulness, ease of use, and satisfaction.

Why it matters: Marginally relevant to learning sciences and skill acquisition (spatial/visual thinking), but the domain is narrowly math education with no HCI, creativity, or behavioral science angle.

Method: ADDIE instructional design model with Content Validity Index (CVI) scoring from expert validation and descriptive statistics from 50 students.