Increasing inequality and negative effects on job quality
Advances in LMs and the language technologies based on them could lead to the automation of tasks that are currently done by paid human workers, such as responding to customer-service queries, with negative effects on employment [3, 192].
ENTITY
1 - Human
INTENT
3 - Other
TIMING
2 - Post-deployment
Risk ID
mit228
Domain lineage
6. Socioeconomic and Environmental
6.2 > Increased inequality and decline in employment quality
Mitigation strategy
1. Establish comprehensive human capital investment mandates, compelling and incentivizing organizations to provide proactive **upskilling and reskilling programs** focused on developing competencies complementary to AI (e.g., critical thinking, complex problem-solving, and socio-emotional skills). Concurrently, implement **portable social safety nets** (e.g., flexible, tax-deferred worker retraining accounts and portable health/retirement benefits) to mitigate economic precarity for workers navigating task-level automation and job transition. 2. Enact targeted **labor market and regulatory reforms** designed to restore worker power and ensure the equitable distribution of productivity gains from AI. This includes clarifying and enforcing independent contractor classification rules to ensure proper employee benefits, loosening unnecessary occupational licensing restrictions to facilitate mobility, and actively promoting mechanisms for **collective bargaining** to afford workers a voice in AI deployment decisions and the adoption of alternative work models (e.g., four-day workweeks) as a form of profit sharing. 3. Require organizations to engage in ethical and transparent **workforce planning and job redesign**, strategically reallocating labor from automatable tasks to high-value functions requiring human judgment. This must be coupled with mandatory **governance and transparency protocols**, including clear communication about AI's impact on specific roles, institutionalizing independent human oversight for all AI-driven employment and personnel decisions, and establishing fair, bias-free metrics for evaluating performance in an augmented workplace.
ADDITIONAL EVIDENCE
A greater risk may be that, among new jobs created, the number of highly-paid “frontier” jobs (e.g. technology devel- opment) is relatively low, compared to the number of “last-mile” low-income jobs (e.g. moderating content in a LM application) [10]. In this scenario, LMs may exacerbate income inequality and as- sociated harms, such as political polarisation, even if they do not significantly affect overall unemployment rates [86, 127].