Recent advances in generative artificial intelligence have significantly improved the fluency and responsiveness of human–AI interaction. However, purely generative approaches remain insufficient for enabling dynamic and flexible interaction in real-world environments, which are inherently uncertain, evolving, and context-dependent. Effective interaction over extended periods requires structured representations, different forms of cognitive principles, and mechanisms for adaptive control. These principles and mechanisms are relevant for good and fluent Human-AI Interaction. Humans navigate such complex uncertain environments by leveraging mental models and a sense of control, continuously adjusting their actions based on perceived affordances and situational constraints.

This talk argues that cognitive architectures, such as ACT-R (Adaptive Control of Thought—Rational), provide a critical foundation for modeling these capabilities in AI systems. By integrating symbolic and subsymbolic representations, ACT-R enables the development of systems that can maintain and update mental models, reason over context, and adapt behavior in a goal-directed manner. Such architectures support continuity, coherence, and interpretability in interaction, particularly over longer time horizons as well as anticipation of the individual in a situation. We propose that combining generative AI with cognitive architectures offers a promising path toward more robust, human-aligned AI systems capable of meaningful, sustained collaboration in dynamic environments.

References

Kahl, S., Wiese, S., Russwinkel, N., & Kopp, S. (2022). Towards autonomous artificial agents with an active self: modeling sense of control in situated action. Cognitive Systems Research, 72, 50-62. https://doi.org/10.1016/j.cogsys.2021.11.005.

Prezenski, S., Brechmann, A., Wolff, S., & Russwinkel, N. (2017). A cognitive modeling approach to strategy formation in dynamic decision making. Frontiers in Psychology, 8, 1335. https://doi.org/10.3389/fpsyg.2017.01335.

Scharfe, M. S. L., Zeeb, K., & Russwinkel, N. (2020). The impact of situational complexity and familiarity on takeover quality in uncritical highly automated driving scenarios. Information, 11(2), 115. https://doi.org/10.3390/info11020115.

Heinrich, N.W., Österdiekhoff, A., Kopp, S., & Russwinkel,N. (2025). Using Eye Movements to Understand Sense of Control in Situated Action, Cognitive Science , vol. 49, pp. e70154, 2025. Wiley Periodicals LLC, https://doi.org/10.1111/cogs.70154.

Klaproth, O. W., Halbrügge, M., Krol, L. R., Vernaleken, C., Zander, T. O., & Russwinkel, N. (2020). A neuroadaptive cognitive model for dealing with uncertainty in tracing pilots' cognitive state. Topics in cognitive science, 12(3), 1012-1029. https://doi.org/10.1111/tops.12515.