Source-linked AI summary
Meeting the Coming Wave: The Emerging Politics of AI and Work across 33 Parliaments
Juliana Chueri, Petter Törnberg
TL;DR
Comparative politics lacks evidence on how parties construct AI’s consequences for work before those consequences are settled. Using 1,514,950 speeches from 33 parliaments, the paper finds that parliamentary conflict centers on enabling versus governing adoption rather than compensating disruption.
Problem
Comparative politics has yet to map how parties construct AI and work politically while its labor-market consequences remain unsettled.
Method
The study analyzes 1,514,950 parliamentary speeches from 33 national parliaments using multilingual retrieval, validated language-model coding, and hierarchical models.
Results
Enablement and investment account for 55.2% of response-frame mentions, versus 21.8% regulation and restriction, 20.6% training, and 2.3% compensation.
Takeaways & Limitations
The emerging AI conflict concerns whether technological change should be enabled or governed, with most parties favoring enablement and the radical left clearest on restriction.
Takeaways & Limitations
Parliamentary speech does not directly measure enacted policy, and the early-period structure may change rapidly as AI’s labor-market effects become visible.
Abstract
from arXiv · showhide
A new politics of artificial intelligence and work is taking shape across party systems, but comparative politics has yet to map it. Using 1,514,950 parliamentary speeches from 33 parliaments (2023-2026), we show this politics follows a different logic than political economy expects. Research anticipates that technological disruption generates demands for compensation; instead, compensation accounts for just 2.3% of response-frame mentions, while enablement and investment dominate (55.2%), regulation and restriction follow (21.8%), and training (20.6%) appears at similar rates across families. Parties disagree instead over what AI means for work and how far this technology should be restrained. The mainstream and radical right support unrestricted enablement; the left is critical but divided on remedy. Social democrats stay adoption-oriented; greens split evenly. The radical left is the clearest force for restriction. The AI conflict thus concerns not compensation after disruption, but whether politics should enable technological change or govern its trajectory.
1 Introduction
AI and work have become an emerging arena of party conflict, but existing political-economy research has focused mainly on compensation after disruption. Across 33 parliaments, the emerging politics instead centers on whether AI adoption should be enabled, regulated, or restricted.
- AI is entering occupations and workplaces while raising concerns about job loss, deskilling, surveillance, algorithmic management, bargaining power, and productivity gains.
- Existing automation research expects political responses centered on compensation, social insurance, redistribution, or retraining after labor-market disruption.
- Parties may contest the pace, direction, restriction, and regulation of technological adoption itself, not only compensation and adaptation.
- The paper analyzes 1,514,950 substantive speeches from 33 national parliaments between 2023 and 2026 using multilingual retrieval, validated language-model coding, and hierarchical models.
- 55.2% of broad response-frame mentions concern enablement and investment, compared with 21.8% regulation and restriction, 20.6% training, and 2.3% compensation.
- Most parties treat AI as an opportunity to enable and manage, while the radical left most clearly contests adoption through regulation and restriction.
2 Shaping technological change: AI and the politics of work
The paper reframes technological politics as contestation over how change is governed before its consequences are settled. It distinguishes what parties diagnose AI to mean for work from the responses they attach to those diagnoses.
- Traditional labor-market politics begins with technological displacement, wage pressure, or occupational hollowing and then debates compensation, social protection, redistribution, or retraining.
- AI-related citizen preferences extend beyond compensation to the pace of technological change, regulation, work protection, digitalization, and investment in adaptation and skills.
- Because institutions, regulation, managerial strategies, and labor power shape technological effects, public authority may accelerate or constrain disruption.
- These responses differ in purpose: compensation absorbs losses, training adapts workers, enablement accelerates adoption, and regulation shapes deployment’s pace and terms.
- Diagnostic frames define AI as a threat or opportunity, while response frames identify regulation and restriction, enablement and investment, training, or compensation.
3 Data and methods
The study constructs a multilingual parliamentary-speech corpus and measures AI-work debate through a recall-oriented retrieval stage followed by validated classification. It then models diagnostic and response frames across party families, while interpreting government-party differences descriptively.
- The corpus contains 1,514,950 substantive interventions from 33 national parliaments between January 2023 and April 2026.
- The non-probability sample spans wealthy and middle-income economies, parliamentary and presidential systems, and diverse party systems using recoverable official speech records.
- A two-stage strategy uses broad multilingual dictionaries for high-recall retrieval, producing 19,411 candidates, then classifies AI-work relevance.
- AI-work relevance covers explicit or broader connections between AI and employment, tasks, wages, skills, job quality, workplace control, labor institutions, or worker protection.
- The coding scheme separates threat and opportunity diagnoses from regulation, enablement, training, and compensation responses, allowing multi-label assignments.
- Against 538 lead-coder observations, AI-work relevance reaches 97.8% accuracy, κ = 0.947, 96.9% precision, and 95.7% recall; diagnostic and response micro-F1 are 0.893 and 0.878.
- Government-versus-opposition estimates are descriptive associations after adjustment, not causal effects of entering government.
4 Findings
AI-work politics is expanding unevenly, but party families differ less in attention than in how they diagnose AI and choose responses. The emerging conflict runs from opportunity-oriented enablement toward threat-oriented regulation, while compensation remains marginal.
- 4.1 AI-work politics is emerging, but unevenly: AI candidate speeches rose from 1.06% of substantive speech in 2023 to 1.70% in 2026, while AI-work connections rose from 18.0% to 33.8%.Even in 2026, AI-work speech remained well below one percent of substantive parliamentary debate.
- 4.1 AI-work politics is emerging, but unevenly: Taiwan and Singapore had the highest AI-work salience, while country profiles varied from threat-oriented Spain, Chile, Belgium, and Italy to opportunity-oriented United States, Canada, Taiwan, and Germany.Enablement and investment was the largest primary response in most countries, while compensation was marginal almost everywhere.
- 4.2 Party conflict begins with diagnosis: Party-family differences in AI-work attention were modest, with overlapping uncertainty intervals and no clear issue owner.The major differences appeared in how parties constructed AI once they addressed its consequences for work.
- 4.2 Party conflict begins with diagnosis: Roughly nine in ten radical-left speeches framed AI as a threat, whereas liberal, conservative, Christian-democratic, and radical-right parties invoked opportunity in roughly 80–86% of diagnostically clear speeches.Greens leaned toward threat, while social democrats made threat and opportunity visible at similar rates; radical-right opportunity framing resembled the mainstream right.
- 4.3 An asymmetric conflict over technological direction: Threat-diagnosing speeches leaned toward regulation and restriction, while opportunity-diagnosing speeches overwhelmingly favored enablement and investment.Social democrats combined a regulatory threat stream with an adoption-oriented opportunity stream.
- 4.3 An asymmetric conflict over technological direction: 55.2% of broad response-frame mentions concerned enablement and investment, compared with 21.8% regulation and restriction, 20.6% training, and 2.3% compensation.Enablement dominated the right and remained largest among social democrats; regulation and restriction dominated the radical left and stood roughly level with enablement among greens.
- 4.3 An asymmetric conflict over technological direction: Party families aligned along an enablement–regulation axis, from treating AI primarily as an opportunity to be enabled to treating it primarily as a threat to be regulated.Compensation remained rare despite a broad definition, while training appeared at broadly similar rates across most party families.
- 4.5 Exploratory analysis: The broad response ordering remained robust under a narrower AI-work definition and after excluding Singapore and Taiwan.Government parties were somewhat more opportunity- and enablement-oriented, whereas opposition parties were more threat- and regulation-oriented.
5 Discussion and conclusion
Across 33 parliaments, AI-work politics is organized less around compensation after disruption than around whether technological change should be enabled or governed. Most parties favor adoption-oriented management, while the radical left and, more ambiguously, the greens make AI’s trajectory politically contestable.
- The enablement–regulation axis captures the central conflict: whether public authority should accelerate and invest in AI adoption or condition and constrain it.
- Most party families, including the mainstream right, radical right, and social democrats, favor enabling AI and preparing workers, firms, and states for transformation.
- The radical left is the clearest parliamentary force for restriction, while greens combine threat-oriented diagnoses with an even split between enablement and restriction.
- The radical right emphasizes opportunity, adoption, competitiveness, and national capacity rather than politicizing AI as a threat to workers.
- The parliamentary record may weakly represent technological insecurity, but the evidence cannot establish a direct representation gap because demand- and supply-side studies cover different populations, countries, and periods.
- Parliamentary speech measures political framing rather than enacted policy, and the early period studied may change as AI’s labor-market effects stabilize.