Causal AI
Categories: Research using linked data, ADR Scotland
21 September 2026
Causal analysis combined with Artificial Intelligence (AI) tools results in a powerful combination that is Causal AI. These advanced methods can leverage large-scale heterogeneous datasets to gain causal insights, such as effects and interventions.
The systematic linking of health and administrative data enables the application of such advanced methods as the data capture interventions over time and heterogeneous circumstances of individuals.
Research focus
Causal analysis can be performed on three granularity levels:
- systemic (population effects and relationships in understanding policy changes),
- groups (households and/or data-driven groupings), and
- individual characteristics (personalised effects and interventions).
Causal AI methods enable various types of analyses:
- Static and over-time data (snapshots and longitudinal)
- Identification of causal drivers (causal discovery)
- Estimation of causal effects (causal inference)
- Recommendation of interventions (e.g. policy change)
Project aims
The aim is to identify opportunities for exploring the application of causal AI methods in the National Safe Haven.
Currently, two research themes are providing immediate opportunities for exploring the application of causal AI methods:
Mental health distress research
In collaboration with Police Scotland, we aim to apply causal AI to data related to police response (e.g., system failure) to incidents of mental health distress.
We aim to inform policing procedure changes that may improve response pathways for frontline public services.
This is possible as the linked health and administrative police data can be used to identify clusters (groups) of individual experiences and outcomes. Most importantly, due to the temporal sequencing of health events and police response events. The RAPID dataset enables the application of causal AI to explore the complex pathways between benefit receipt, employment outcomes, and health for individuals with long term health conditions or disabilities.
The data
- Health and administrative police data
- RAPID dataset
Research team
Professor Sotos Tsaftaris, Professor Chris Dibben, Dr Theoni Massara, Dr Kurt Butler, and Dr Damian Machlanski
Publications, outputs and media
- Sept 2026 - What is Causal AI - view CHAI Hub video for an explanation
Categories: Research using linked data, ADR Scotland