Reflecting on AI in society

Artificial intelligence is part of our everyday life, a component of technical innovations and increasingly becoming part of education. Intelligent systems are often assumed to make objective decisions, but numerous examples demonstrate how such systems take over our stereotypes and (un-)conscious biases. We inspect emerging ethical and societal implications within the Interchange Forum for Reflecting on Intelligent Systems (IRIS) and the “Platform for Reflection” of the DFG-funded Cluster of Excellence “Data-Integrated Simulation Science” (EXC 2075). Ongoing projects in this area relate to understanding cognitive mechanisms underneath reflection, AI-generated biases, and human trust.

Managing AI-generated biases (2022 – present)
In an ongoing collaboration, we inspect cognitive mechanisms underneath managing AI-generated bias. We build on cognitive evidence from decision making research to develop computational models that can predict determinants of behavioral change. Future steps will involve applying these insights to focused training interventions.

Cognitive mechanisms of trust in AI (2022 – present)
A dissertation project affiliated with the SimTech Graduate Academy focuses on cognitive factors that contribute to or diminish trust in AI. Here, we particularly consider aspects such as algorithm aversion / appreciation, cognitive dissonance, and (un-)conscious biases. In a first step, we inspected effects of discursive interactions with a chatbot as part of a public engagement project. Subsequently, Bayesian modeling will serve as means to formally decompose and model trust in AI.

Mechanisms of reflective learning (2019 – present)
Reflective learning relates to human’s ability to introspectively examine their own learning process. This happens by sequential learning episodes that iteratively re-evaluate the trustworthiness of acquired experiences and information gained in each learning step to face potential future difficulties more resiliently. Applying Socratic questions in the form of metacognitive prompts has already been proven to result in a more optimized investment of limited cognitive resources and subsequently more sustainable performance results. We investigate cognitive mechanisms underneath reflective learning with prompts (among others) particularly related to interacting with intelligent systems.

Related publications

  • Chen, S., & Wirzberger, M. (in preparation). Reflection in AI: explore conflict-exploit comfort dilemma.
  • Becker, F., Wirzberger, M., Pammer-Schindler, V., Srinivas, S., & Lieder, F. (under review). Systematic metacognitive refletion helps people discover far-sighted decision strategies: a process-tracing experiment. Judgement and Decision Making. [PREPRINT]
  • Berberena, T., & Wirzberger, M. (2021). Bringing Thiagi to the classroom: Reducing stereotype-threat by promoting reflection in CRT. In EARLI 2021 – Online. [PDF]
  • Berberena, T., & Wirzberger, M. (2021). Unveiling unconscious biases and stereotypes in students: The necessity of self-reflection in Higher Education. In T. Fitch, C. Lamm, H. Leder, & K. Teßmar-Raible (Eds.), Proceedings of the 43rdAnnual Meeting of the Cognitive Science Society (p. 3488). Cognitive Science Society. [ABSTRACT]