RIT will host a CHAI Seminar plenary talk by Dr. Paulo Shakarian, KG Tan Endowed Professor of Artificial Intelligence at Syracuse University, on Monday, September 21, 2026, 12:00-1:00 PM in the Simone Center 1600 Atrium. The talk, titled "Improving LLM Performance with Metacognition," addresses how recent advances in large language models have primarily stemmed from scaling laws, where correctness is defined in terms of model fit to historical data, but general use-cases cause an objective function mismatch because the LLM was trained with one objective while being applied to a problem with a different one. LLMs can fail because they may not be employed in a constrained manner, may provide results inconsistent with domain knowledge, and may exhibit common failure modes in certain use cases. Shakarian presents three approaches: using abductive inference to constrain LLM use for targeted text reframing; showing how domain knowledge can inform an LLM cybersecurity agent by enforcing logical constraints; and examining how psychology-inspired metacognitive techniques can reduce errors in a small language model, allowing for local queries and limiting use of more expensive LLMs. Shakarian directs the Leibniz Lab and has made notable contributions in logic programming, neurosymbolic AI, security, and data mining.