Societal Impact Assessment of AI-Driven Automation Systems

Abstract

AI-driven automation systems are transforming labour markets, organisational structures, and social arrangements at an unprecedented pace, with societal impacts that extend far beyond the immediate efficiency gains to individual deploying organisations. Comprehensive societal impact assessment (SIA) for AI automation -- evaluating impacts on employment, wage distribution, occupational identity, community resilience, democratic participation, and generational opportunity -- is urgently needed but methodologically underdeveloped: most existing impact assessments focus on productivity and efficiency metrics, treating labour market disruption as an externality rather than a primary impact dimension. This paper proposes the AI Automation Societal Impact Assessment Framework (AASIAC), a systematic methodology for comprehensive SIA of AI automation across six impact domains -- employment and labour market, wage distribution, occupational identity and dignity, community and regional economic resilience, democratic and political participation, and intergenerational opportunity -- applied to four automation deployment scenarios: manufacturing AI, service sector AI, professional knowledge work AI, and public sector AI. AASIAC introduces the Societal Impact Severity Score (SISS) and evaluates each scenario against six impact domains using structured expert elicitation (20 economists, sociologists, and political scientists), quantitative economic modelling, and community impact case studies. Key results: professional knowledge work AI (LLM-driven automation) achieves the highest SISS (0.848) due to breadth of occupational disruption and concentration in urban knowledge economy centres; wage distribution impact is the most severe single domain across all scenarios; community resilience is the most neglected impact dimension in current policy analysis. The framework provides SIA methodology guidance for policymakers, technology deployers, and affected communities.