Thought experiments have a long and fascinating history. As noted in the Stanford Encyclopedia of Philosophy, they were used by ancient Greek philosophers and later in both the natural and social sciences. However, the use of AI to help create and develop thought experiments is a relatively new development which remains largely unexplored.
When prompted, ChatGPT has outlined the following potential positive and negative consequences of using AI in thought experiments.
Positive contributions of using AI
- Enhances plausibility and depth – draws on large datasets to create more realistic and internally consistent scenarios
- Enables rapid iteration of scenarios – allows multiple versions of a thought experiment to be tested under different assumptions
- Improves transparency of assumptions – makes it easier to vary and expose key drivers shaping outcomes
- Supports modelling of complex systems – captures interactions between institutions, individuals, and technologies more effectively
- Strengthens analytical rigour – moves thought experiments closer to structured, evidence-informed analysis
- Improves communication and accessibility – generates narratives, case studies, and visualisations that make ideas more engaging for policymakers
Negative risks of using AI
- Bias toward existing patterns – tends to reproduce current trends rather than imagine genuinely disruptive futures
- Reduced conceptual originality – may limit creative, counterintuitive thinking by steering users toward familiar ideas
- Risk of “synthetic plausibility” – produces convincing but weakly reasoned scenarios that lack theoretical clarity
- False sense of precision and certainty – can make speculative scenarios appear more evidence-based or predictive than they are
- Hidden assumptions and bias – AI-generated outputs may embed biases that are difficult to identify or challenge
- Over-complexity – adds excessive detail, potentially obscuring core arguments and causal relationships
The potential positive contributions and negative risks of using AI will be taken into account during this project and commented on after the project has finished.
HE2050 