AI-Simulated Expert Panels for Socio-Technical Scenarios and Decision Guidance

Abstract

Socio-technical scenarios for net-zero and other transformation pathways combine qualitative storylines with quantitative models, embedding them in plausible societal contexts for model assessment. Conventional scenario generation is resource-intensive, can be limited in internal consistency and diversity of expert and stakeholder perspectives, and is rarely stress-tested. This paper introduces a synthetic, AI-based expert panel to address these bottlenecks. An AI model first simulates domain experts who agree on descriptors, states, and their interactions. A probabilistic Cross-Impact Balance analysis then generates internally consistent pathways, using stochastic shocks to assess robustness and pathway diversity. An AI stakeholder panel uses multi-criteria decision analysis to select a preferred pathway; an AI expert panel translates it into model-ready quantitative inputs. Although scalable and applicable to any other country or region, the framework is applied to Germany's energy transition as a proof of concept, and offers an alternative and/or supplement to scenario generation. Furthermore, it enables Virtual AI-Led Decision Laboratories for exploratory policy stress-testing and provides an approach for rapid, structured expert elicitation and decision support in other domains.

Figures (2)

• Figure 1: Cross-impact matrix (CIM) as descriptor-level impact networks and the effect of the two stochastic shock mechanisms (structural and dynamic). Panels (a)-(c) share the same node-size scale (total outgoing impact strength). a, Base: workshop-agreed CIM; edge width is link strength (top 25% by weight). b, Structural shock (standard deviation 0.30): same layout; edge width reflects shock magnitude, node size is total outgoing impact in the shocked CIM. c, Shocked - base: edge width reflects change in link weights (shocked CIM minus base); node size as in (a). d, Dynamic shocks (one period, 2040): node size and colour indicate mean absolute dynamic shock intensity per descriptor (scale not comparable to (a)-(c)). Data from synthetic AI workshops, CIM computed using PyCIB PyCIB.
• Figure 2: Selected results of the AI-panel elicitation for socio-technical energy scenarios for Germany's energy transition. a, CIB: Decarbonisation outcome (all three states). b, CIB: one state each for Hydrogen role, Policy stringency, Renewables deployment. In (a)-(b), central lines are ensemble shares (the fraction of 10,000 Monte Carlo runs in which the descriptor is in that state at that time); bands are Wilson score confidence intervals (with a visual minimum-width fan). c, MCDA: Decarbonisation outcome for four candidate pathways; selected pathway highlighted, others and a CIB sample for context. d, MCDA: score matrix (pathways $\times$ criteria), scores 1-5. e, Quantification: CO$_2$ emissions and carbon price. f, Technology cost index and real WACC. In (e)-(f), central lines are trajectories as $\Delta$% from 2025; bands are expert-elicited uncertainty ranges. g, Public acceptance and policy stringency (state over time); central trajectory smoothed (cubic spline), bands as in (e)-(f) mapped to state space. Underlying data are period-wise and step-wise; within-state variation not represented. Data from synthetic AI workshops.