Dialogical Sequence Analysis (DSA) is a method for understanding how people position themselves in speech — developed by Professor Mikael Leiman over 30 years of clinical research. It is not a coding scheme but a way of listening.
The DSA-Intelligence project, supported by a Thinking Machines Tinker Research Grant, explores whether AI can be trained to support this kind of deep clinical listening. The goal is not to replace the human analyst but to assist researchers and clinicians in their work.
The core challenge is fascinating: DSA is fundamentally about hearing semantic positions — the networks of signs a person is connected to, their stance toward those objects, and how the addressee shapes what can be said. Teaching a machine to hear these patterns requires moving beyond surface-level text analysis.
The project combines psychotherapy process research expertise with modern AI methods. Early results suggest that with carefully curated training data and theoretically grounded approaches, AI can learn to identify some of the patterns that trained analysts recognize.