Jakub Rada, Viliam Lisý AI Center, Department of Rehabilitation Science et al. (5 authors) • Semantic Scholar
Finding: LLMs play suboptimally in sequential decision-making games (tic-tac-toe, Connect Four) even against simple MCTS opponents, and an experience memory framework with post-game reflection improves performance without weight modification.
Why it matters: Tangentially relevant to AI capabilities and games research, but the focus is narrow AI benchmarking with minimal connection to HCI, learning sciences, or game design as the professor teaches them.
Method: Used obfuscation tests (rewriting surface form while preserving game tree structure) to isolate memorization from actual reasoning ability.
Alli Abdurrazaq, Omar D Camara, Saidu Conteh et al. (4 authors) • Semantic Scholar
Finding: AI-assisted multilingual instruction (CTCA 2.0 framework) significantly improved student attitudes toward learning computing concepts but did not significantly improve academic performance compared to traditional lectures.
Why it matters: Weak relevance—touches on AI in education and learning engagement, but focuses narrowly on computing pedagogy and multilingual instruction rather than HCI, creativity, teams, or behavioral science.
⚠️ Small sample (n=75) | Single study (not replicated)