Participatory Technology Assessment

In the AI Story

The Danish Board of Technology's consensus conferences, beginning in the mid-1980s, established the founding model. Fifteen ordinary citizens — selected for demographic diversity and lack of prior expertise — were given briefing materials, access to expert witnesses they could question at length, and structured time for deliberation. The resulting reports were presented to parliament and influenced subsequent legislation. The model demonstrated that citizens without technical training could deliberate productively on complex technologies, identify consequences experts had not anticipated, and produce recommendations that carried democratic legitimacy precisely because they emerged from a process not dominated by any particular expertise or interest.

Jasanoff's scholarship on PTA emphasizes both its successes and its limitations. The successes are real: participatory processes produce governance that is more responsive to public values, more attentive to distributional consequences, and more legitimate in the eyes of those governed. The limitations are equally real: PTA is slow, expensive, and requires institutional infrastructure (skilled facilitation, adequate information, protected time) that is difficult to maintain. The temporal mismatch between deliberative processes (operating on timescales of months) and AI development (operating on timescales of weeks) creates a structural challenge that no existing PTA model has solved.

The Trivandrum training that Segal documents exemplifies governance without participation. Twenty engineers experienced a professional transformation — job descriptions changed, specialist roles dissolved, expertise was redefined — within a week. The transformation was impressive and, from a capability perspective, beneficial. But the engineers were not consulted about whether the transformation was desirable. Their consent was assumed on the basis that the capability gain was self-evidently good. A participatory framework would have asked different questions: What do you need from this transformation? What do you fear losing? What support would help you navigate it? What would it mean for you to have voice in how your professional identity is restructured?

Jasanoff's framework identifies three institutional models with potential applicability to AI governance. Workers' co-determination, practiced in Nordic countries, gives employees institutional standing in decisions about workplace technology deployment. When a Swedish company adopts AI tools, affected workers have legal rights to consultation and negotiation before implementation. Citizens' assemblies, demonstrated in Ireland's abortion referendum and France's climate convention, show that randomly selected citizens can deliberate on divisive issues and produce democratically legitimate recommendations. Community benefit agreements, common in infrastructure development, make communities active negotiators rather than passive recipients of technological change. Each model is imperfect, each faces scaling challenges, and each requires institutional capacity that is scarce. But each embodies a principle essential to Jasanoff's vision: that the people who bear the costs of technological transitions deserve institutional voice in managing those transitions.

Origin

Participatory technology assessment originated in Denmark in the early 1980s as a response to public distrust of expert-dominated decision-making about nuclear energy and biotechnology. The Danish Board of Technology, established in 1986, institutionalized the consensus conference model and influenced similar institutions across Europe — the Dutch Rathenau Institute, the Norwegian Board of Technology, the Swiss Centre for Technology Assessment. The practice has been studied extensively by scholars including Richard Sclove, Daniel Lee Kleinman, and Jasanoff herself, whose comparative work shows that PTA takes different forms in different political cultures and that its success depends on institutional design that matches cultural context.

Key Ideas

Citizens possess relevant knowledge. The experiential knowledge of people living with a technology's consequences is epistemically essential to governance — not inferior to expert knowledge but complementary, addressing questions experts cannot answer from technical analysis alone.

Participation produces legitimacy. Governance decisions gain democratic legitimacy through the inclusion of affected voices with genuine authority, not merely consultative status — transforming people from subjects of governance into participants in it.

Multiple PTA models exist, none adequate to AI. Consensus conferences, stakeholder consultations, co-determination frameworks — each captures some needed knowledge but faces temporal and scalar mismatches when applied to technologies developing as rapidly as AI.

Institutional design is the challenge. Effective PTA requires standing mechanisms, adequate resourcing, skilled facilitation, protected time, and cultural change within governance institutions to recognize experiential knowledge as evidence rather than anecdote.