Democratic technology assessment is the institutional construction that brings the voices of affected communities into technology governance with genuine authority rather than merely consultative status. It addresses a fundamental democratic deficit: AI is experienced locally (in specific workplaces, schools, neighborhoods) but governed nationally or internationally at abstraction levels that exclude those most directly affected from meaningful participation. Assessment happens at community level through structured processes—deliberative forums, participatory panels, digital platforms for collective input—where parents, teachers, workers, students examine AI deployment's effects in their particular contexts, develop recommendations grounded in experiential knowledge, and communicate those recommendations to governance bodies with institutional weight sufficient to affect outcomes. This is not populism (which assumes mass preference should determine all choices) but recognition that governance knowledge is distributed: people living inside arrangements possess insight into their effects that external designers cannot access, and democratic governance is enriched rather than undermined by including this knowledge in institutional design and revision.
Current technology governance operates through a vertical model: expert bodies (regulatory agencies, legislative committees, corporate boards) make decisions based on technical knowledge and aggregate data, then impose those decisions on populations whose lived experience of the technology is excluded from the decision-making process. This model was functional when technological change was slow enough that representative democracy's filtering mechanisms (elections, public comment, media discourse) could adequately channel citizen input. It is dysfunctional when technological transformation proceeds faster than these mechanisms operate and when the effects are so locally specific that aggregate data cannot capture the relevant variation.
Democratic technology assessment constructs horizontal channels supplementing vertical governance. Community-level assessment panels—stratified for demographic representativeness, provided with technical briefings and facilitation, given structured time for deliberation—examine AI deployment in their specific contexts: this school's use of AI tutoring systems, this hospital's AI diagnostic tools, this factory's AI-augmented production systems, this municipal service's AI decision-making. The assessment is not mere opinion-gathering but structured evaluation: What effects are observable? What values are at stake? What trade-offs are being made? What alternatives might better serve the community's purposes? The results—grounded in experiential knowledge experts cannot access—inform governance bodies with genuine institutional weight.
The institutional architecture must solve several design challenges. Accessibility: participation must be genuinely open to affected populations, not merely those with free time, technical literacy, and institutional confidence. Quality: deliberation must be informed enough to produce useful governance knowledge, requiring technical briefings, structured processes, facilitation supporting careful evaluation rather than mere preference expression. Authority: assessments must influence actual decisions rather than being filed as "community input" that governance bodies are free to ignore—binding recommendations in some domains, weighted input in others, but always with institutional mechanisms ensuring community voice affects outcomes. Speed: the assessment process must operate at tempo adequate to technological change, producing recommendations while deployment decisions remain open rather than arriving after arrangements have crystallized.
Precedents exist but require extension for AI's pace and scope. Brazilian participatory budgeting demonstrates community-level deliberation with binding authority over resource allocation. Irish citizens' assemblies demonstrate sortition-based bodies producing recommendations on complex issues that legislative bodies then implement. Scandinavian co-determination demonstrates worker participation in technology deployment decisions within firms. Each precedent confirms that democratic technology assessment is constructible rather than utopian. What the AI transition demands is their synthesis and acceleration: community-level assessment operating continuously rather than episodically, at national and international scales, with authority extending beyond resource allocation to the fundamental design of institutional arrangements governing AI deployment.
Technology assessment as a governance practice emerged in the 1970s with the U.S. Office of Technology Assessment and European parliamentary TA offices. But these were expert-driven rather than democratic—analysts evaluating technologies on behalf of legislatures rather than communities evaluating technologies affecting them directly. Democratic TA emerged from participatory democracy movements (Brazilian experiments, Scandinavian workplace democracy) and was theorized by Archon Fung, Erik Olin Wright, and other deliberative democracy scholars. Unger's contribution is integrating democratic TA into the experimentalist governance framework as essential component of high-energy democracy adequate to rapid technological transformation.
Horizontal supplementing vertical. Community-level assessment channels providing experiential knowledge to complement expert analysis—both required for adequate governance, neither sufficient alone.
Genuine authority not consultation. Assessment results must influence actual decisions with institutional weight rather than being filed as input that governance bodies are free to ignore—binding power in some domains, weighted influence in others.
Structured deliberation not polling. Informed evaluation through briefings, facilitation, structured comparison of alternatives rather than mere preference expression or opinion aggregation.
Pace adequate to transformation. Assessment operating continuously rather than episodically, producing recommendations while deployment decisions remain open rather than arriving after arrangements crystallize into naturalized necessity.