Finding: This systematic review maps generative AI applications in manufacturing and industrial contexts, finding that while 72% of studies report empirical evaluation, only 0.8% show operational deployment evidence.
Why it matters: The gap between publication activity and deployment maturity is mildly relevant to AI's impact on work, but the focus is narrowly on manufacturing/engineering rather than HCI, collaboration, learning, or behavior change.
Modupe Ruth Omofoye, O. Awodele, Oluwatosin Amoke Fabiyi et al. (4 authors) • Semantic Scholar
Finding: A structured literature review proposes a four-layer "Agent-Driven Inclusive AR" (ADI-AR) framework integrating human-centered AI, adaptive interfaces, and pedagogical scaffolding to make augmented reality educational platforms accessible and inclusive, with approval-based monitoring to reduce automation bias.
Why it matters: Directly relevant to HCI/UX design in educational technology, AI's impact on learning, and behavior change mechanisms like cognitive overload management and automation bias—all core teaching areas for this professor.
Method: Structured literature review across IEEE Xplore, ACM Digital Library, ScienceDirect, SpringerLink, and Google Scholar (2005–2025).
Jargon: *Automation bias* = over-reliance on automated system suggestions; *pedagogical scaffolding* = temporary support structures that help learners reach higher skill levels; *ADI-AR* = the authors' proposed four-layer framework combining sensing, AI agents, adaptive UI, and human oversight.
Sura Genc, Marc Wittmann, Bigna Lenggenhager et al. (4 authors) • PsyArXiv
Finding: Continuous real-time measurement of flow states (using a foot pedal + HRV) captures temporal dynamics that retrospective self-reports miss, with game performance as the strongest predictor of flow.
Why it matters: Flow measurement methodology is directly relevant to game design, gamification, engagement mechanics, and behavior change design—particularly for understanding moment-to-moment user engagement rather than just post-hoc satisfaction.
Method: Participants continuously reported flow via a custom foot pedal during 40-minute VR game sessions while ECG/HRV and performance data were recorded simultaneously; mixed-effects models and cross-correlation analysis examined temporal relationships.
Jargon: *Flow Short Scale (FSS)*: validated retrospective questionnaire measuring absorption and flow. *HRV (Heart Rate Variability)*: variation in time between heartbeats, used as a proxy for autonomic nervous system arousal. *HF-HRV*: high-frequency component of HRV reflecting parasympathetic (calming) activity. *LF/HF ratio*: balance between sympathetic and parasympathetic activation.
⚠️ Small sample (n=35) | No control group mentioned | Single study (not replicated)
Attila Márton Putnoki, Tamás Orosz • Semantic Scholar
Finding: Personalized dashboard design that aligns with cognitive load principles and HCI improves decision-making efficiency and user trust in Decision-Support Systems.
Why it matters: Directly relevant to HCI/UX design and behavioral science of decision-making, with implications for how visual interface design affects cognitive biases and choice quality.
Method: Mixed survey data from professional decision-makers combined with dashboard design examples evaluated for usability and engagement.
Jargon: *Cognitive resonance* — alignment between information presentation and user mental models; *CIS (Cognitive Information Systems)* — systems designed around human cognitive processes rather than raw data display.
Finding: Introduces "scale hacking"—HCI methodological strategies for studying large-scale AI harms without massive samples—applied to studying foundation model impacts on the freelance economy, focusing on disclosure practices and sociotechnical risks.
Why it matters: Directly relevant to AI's impact on work and HCI methodology; the freelance economy context and disclosure norms connect to behavioral science and how AI reshapes human collaboration and labor practices.
Method: Triangulation across multiple methods/data streams, longitudinal participant sampling with varying experience levels, and practice-focused rather than application-focused analysis.
Jargon: *Scale hacking* = methodological workarounds to study population-level phenomena with smaller samples; *foundation models* = large general-purpose AI models (e.g., GPT-4); *sociotechnical harms* = negative outcomes arising from the interaction of social systems and technology.
⚠️ No control group mentioned | Single study (not replicated) | Limited statistical detail in abstract
Finding: 76-84% of students adopted an AI tutor during exam-relevant periods; usage positively correlated with academic performance, and trust/attitudes toward AI did not predict usage (situational demand overrode personal disposition).
Why it matters: The cognitive offloading finding and the disconnect between attitudes and actual AI tool adoption are directly relevant to behavioral science (intention-behavior gap) and AI's impact on learning and skill acquisition.
Method: Longitudinal semester-long study tracking AI tutor usage, academic performance, self-efficacy, and attitudes — though labeled "theoretical" by submitter, it reads as empirical.
Jargon: *Cognitive offloading* — delegating mental tasks to external tools (here, AI), which can reduce skill development if overused; *self-efficacy* — belief in one's own ability to perform a task.
Finding: Proposes a framework for integrating trustworthy GenAI into executive decision-making, emphasizing governance-by-design, explainability, and human-AI collaboration rather than post-hoc governance controls.
Why it matters: Directly relevant to AI impact on work and human-AI collaboration, particularly how organizations should structure AI adoption at leadership levels—useful for teaching AI governance and decision-making.
Method: Theoretical framework development aligned with EU AI Act, NIST AI RMF, and ISO standards—no empirical validation.
Jargon: *Governance-by-design* = embedding oversight rules into system architecture from the start; *Risk-adaptive orchestration* = dynamically adjusting AI behavior based on assessed risk levels; *Continuous assurance* = ongoing monitoring rather than one-time compliance checks.
Robert Šamárek, Kristýna Klimková, Vojtěch Koňařík et al. (4 authors) • Semantic Scholar
Finding: Visitors to an art exhibition featuring human vs. AI-generated works segment into four groups (Traditionalists, Visualists, Skeptics, AI Enthusiasts) based on their perceptions of aesthetic and conceptual value of AI art.
Why it matters: Relevant to AI's impact on creative fields and how people perceive machine creativity vs. human creativity—useful for teaching about AI's role in creative work and human responses to it.
Method: Factor and cluster analysis of visitor questionnaire data from a real gallery exhibition using GPT-4o, DALL-E 3, and other generative models.
Jargon: *Cluster analysis* — statistical method grouping respondents by similar response patterns; *TTS-1* — OpenAI's text-to-speech model.
Finding: A Bayesian computational model of sequential collaboration shows that dependent judgments (where people revise others' estimates) can match or outperform aggregated independent judgments, with social influence and expertise parameters quantitatively recoverable from data.
Why it matters: Directly relevant to collective intelligence and team decision-making research, particularly how social influence and anchoring bias shape group judgment accuracy.
Method: Bayesian parameter estimation applied to four empirical datasets, validated through simulation studies for parameter recovery.
Jargon: *Wisdom-of-crowds*: aggregating many individual estimates improves accuracy; *anchoring*: over-reliance on a reference point (here, a prior contributor's estimate); *cultural consensus theory*: model for estimating shared knowledge and individual competence in a group.
⚠️ No control group mentioned | Single study (not replicated)
Finding: An AI-powered system combining hand gesture recognition (via MediaPipe) and voice commands achieves real-time touchless computer control with low latency across cursor movement, scrolling, and application control.
Why it matters: Touchless interaction is relevant to HCI and accessibility design, and the system has stated applications in educational platforms and assistive technology, but the paper focuses narrowly on implementation/engineering rather than user behavior, UX evaluation, or design implications.
Method: Experimental evaluation under varying operating conditions measuring recognition accuracy, response time, and reliability — but no user study or usability testing reported.
Jargon: *MediaPipe* — Google's open-source ML framework for detecting body landmarks (hands, face, pose) in real time; *hand landmark detection* — identifying key joint positions on a hand from camera input.
⚠️ No control group mentioned | Single study (not replicated) | Limited statistical detail in abstract
Peter Washington, Anoop N. Muniyappa, Chen Liang et al. (10 authors) • Semantic Scholar
Finding: Proposes using generative AI tools to create interactive simulations and visualizations for teaching AI concepts to health profession students without technical backgrounds.
Why it matters: Tangentially relevant to learning sciences and AI's impact on education, but narrowly focused on health professions education instructional design rather than the core HCI, creativity, or behavioral science topics this professor teaches.
Xuewen Sun, Lan Tian, Weidong Zhou et al. (5 authors) • Semantic Scholar
Finding: XAI methods (SHAP, counterfactuals, gradient attribution) are maturing toward human-centered evaluation and domain-sensitive explanations, with a brief mention of XAI's emerging role in education.
Why it matters: Tangentially relevant to AI's impact on work and human-AI collaboration, but this is primarily a technical survey editorial with minimal focus on the professor's core interests.