📚 Research Digest

Finding: AI contamination is systematically distorting online research data through three pathways (full delegation, partial mediation, and spillover), with LLM use reported in nearly one-third of text-rich crowdwork tasks, creating human-appearing responses that embed model-shaped biases and compress attitudinal extremes.

Why it matters: This fundamentally challenges the validity of online behavioral science research and HCI studies that rely on crowdsourced data, requiring new methodological safeguards for studying human decision-making and collaboration.

Method: Synthesizes emerging evidence to create a taxonomy of contamination pathways and vulnerability zones across different research settings.

Jargon: AI-mediated contamination = when AI assistance systematically distorts research responses while appearing human-generated; spillover = indirect contamination effects beyond direct AI use.

Finding: This scoping review of 23 studies demonstrates that video games can successfully replicate canonical EEG findings (like N2/P3 responses, error-related negativities) while providing higher ecological validity than traditional laboratory tasks, supporting their use as a middle ground between experimental control and real-world relevance.

Why it matters: This validates video games as rigorous research tools for studying cognition, attention, and decision-making, directly informing both game design research and behavioral science methodology.

Method: PRISMA-based scoping review systematically mapping video game EEG studies onto canonical ERP and oscillatory components across stimulus-, response-, and feedback-related events.

Jargon: ERP = event-related potentials (brain responses to specific events); N2/P3 = specific neural response components; ecological validity = how well lab findings apply to real-world situations.

Finding: VLAB, a cross-platform virtual laboratory, successfully captures temporally precise behavioral measurements in online research environments, performing as well as traditional lab setups and sometimes better on smartphones due to touchscreen advantages.

Why it matters: This validates a scalable tool for conducting rigorous HCI research and behavioral experiments outside traditional lab settings, enabling broader participant recruitment and more ecologically valid studies of human-computer interaction.

Method: Used a 2×2 experimental design comparing smartphones vs. computers and lab vs. outside locations with 48 participants completing avoidance conditioning tasks requiring precise temporal measurement.

Jargon: Latent timing = implicit learning of time intervals; scalar property = timing precision that scales proportionally with interval duration; avoidance conditioning = learning to respond to avoid negative outcomes.