Action Research was developed in the 1940s by Kurt Lewin: the same thinker whose change model we explored in Course 2. Lewin was dissatisfied with a social science that produced knowledge without producing change. He wanted research that was simultaneously rigorous and useful, that combined the scientist’s commitment to evidence with the practitioner’s commitment to improvement.

His formulation was simple but radical: the best way to understand a social situation is to try to change it.

Action Research is defined by several principles that distinguish it from conventional research:

Cyclical:   Action Research proceeds in cycles of planning, acting, observing, and reflecting. Each cycle produces learning that feeds into the next. It is iterative by design, acknowledging that complex social situations reveal themselves gradually.

Participatory:   Action Research is conducted with people, not on them. The people living the situation:   whether workers, community members, or AI users:   are active participants in the inquiry, not passive subjects of study.

Action–oriented:   the purpose of the research is not to produce a report. It is to improve the situation. Knowledge and action are inseparable.

Reflective:   Action Research requires ongoing critical reflection on what is happening, what is being learned, and what that means for the next cycle of action.

Contextual:   Action Research does not seek universal laws. It seeks deep understanding of a specific situation, in its full complexity and particularity.

These principles make Action Research particularly well-suited to ethical AI challenges, which are: complex, context-dependent, involving multiple stakeholders, and requiring both understanding and practical improvement.

Reflection question: Think of an ethical AI challenge in your organisation. What would it mean to research it in the Action Research sense:   with the people affected, in cycles, oriented toward improvement?