Algorithmic management expands gig work access in India but embeds opaque, unfair practices that undervalue labor; the authors recommend a pragmatic hybrid 'Algorithmic Human Manager' model to pair technological efficiency with human accountability.

Main Finding

Algorithmic management in India’s location-based gig economy (ride‑hailing, food/grocery delivery, home services) creates a hybrid "algorithmic + human" manager: apps automate allocation, performance monitoring, pricing and incentives at scale, while human teams remain responsible for high‑stakes actions (deactivations, dispute resolution). This hybrid governance delivers efficiency and rapid onboarding but also produces opacity, income fragility, perceived manipulation of incentives, arbitrary penalties/deactivations, and inadequate grievance redressal — undermining worker fairness, dignity and well‑being. A pragmatic “middle path” of pro‑innovation guardrails (algorithmic transparency, human‑in‑the‑loop for critical decisions, participatory design, data infrastructure for workers) is recommended.

Key Points

• Worker experience and perceptions

  • Gig work often chosen as a fallback or necessity rather than preference; attracts those with limited alternatives.
  • Workers primarily interact with the platform app and view it more as a tool of control than empowerment.
  • Most workers are unfamiliar with the term “AI” but can describe its perceived effects (allocation, monitoring, incentives).
  • Chat support is the most used grievance channel but fails on complex, high‑stakes issues (penalties, deactivations, payouts); workers want time‑bound escalation to humans.
  • Workers generally believe humans still make final decisions behind the app; opacity around allocation, surge, penalties and ratings causes distrust.

• Platform practices and technology

  • Digitised recruitment and onboarding (targeted ads, data pipelines, even GenAI) drastically shorten time‑to‑work; inclusion aids (speech‑to‑text) help but digital literacy and intermediaries create new frictions/risks.
  • Algorithmic management tools used: location tracking, selfie verification, dynamic pricing/surge, performance/rating systems, incentive nudges.
  • “Algorithmic cruelty” reported: perceived manipulation of incentives, unexplained deactivations, unilateral rate/fee changes, income obfuscation, and pervasive digital surveillance.
  • Human‑tech collaboration (warehouses/dark stores) increases throughput but hard‑codes rigid targets that ignore exogenous disruptions (weather, traffic), leading to unfair penalties.

• Wages, incentives and financial fragility

  • Some platforms show real‑time earnings and surge logic, but many workers face opaque earnings calculations, arbitrary deductions and unilateral changes that can wipe out weekly income.
  • Financial fragility and stress result from unpredictable pay and penalty regimes.

• Policy and governance landscape

  • Central Code on Social Security (2020) recognises gig/platform workers but implementation (notifications, contributions) remains incomplete.
  • Several states (Rajasthan, Karnataka, Jharkhand, Bihar, Telangana) are advancing laws on transparency, grievance redressal, explainability and welfare.
  • Standards for human‑in‑the‑loop, data rights and grievance procedures are uneven.

• Recommendations (high level)

  • Adopt a pragmatic “middle path”: preserve innovation/efficiency while establishing digitally enabled worker protections.
  • For platforms: transparency about algorithms, human review for high‑stakes automated decisions, participatory design with workers, embed micro‑learning in apps.
  • For ecosystem: create a "Data Gigs" ecosystem to enable alternative digital earning opportunities (annotation, labeling) for gig workers.
  • For governments: build a Unified Worker Interface (UWI) — public digital infrastructure for portable verified ledgers of earnings/work history, portability of reputation, and operationalising social security contributions; enforce time‑bound human escalation and a legal right to explanation.
  • Public repository/catalogue of design and dark patterns to hold platforms accountable and guide remediation.

Data & Methods

• Study produced by Centre for Responsible AI (CeRAI), IIT Madras (September 2025).
• Mixed‑methods approach:
  • Primary qualitative interviews: in‑depth interviews with 16 blue‑collar gig workers (ride‑hailing, food/grocery delivery, home services).
  • Stakeholder interviews: 21 key stakeholders (platform representatives, policymakers, civil society, researchers).
  • Field insight: author experience participating as a platform worker contributed practical exposure.
  • Secondary research and literature review: policy documents (Code on Social Security 2020, state bills), prior studies (NITI Aayog, Fairwork, academic reports) and market estimates on gig workforce size.
• Analytical framing:
  • Normative evaluation using a social justice lens (Rawls’ difference principle): testing whether AI/digital arrangements improve outcomes for the least advantaged gig workers.
  • Thematic coding of interview material to identify recurring patterns: allocation opacity, grievance experiences, onboarding, wages, human‑tech interactions, and suggested mitigations.

Implications for AI Economics

• Labor allocation and matching

  • Algorithmic allocation drives real‑time matching and effectively sets marginal returns to worker time — models of platform labor supply should incorporate algorithmic allocation rules and the transparency/opacity of matching.
  • Opacity creates information asymmetries that affect workers’ labor supply decisions, reservation wages and search behavior; economic models must account for bounded rationality and low AI literacy.

• Wage formation, incentives and risk

  • Dynamic pricing, incentive nudges and unilateral rate changes alter the wage process: earnings are stochastic and subject to non‑transparent adjustments, increasing income volatility and risk premiums demanded by workers.
  • Platforms internalize monitoring/discipline benefits; this can depress wages relative to non‑platform markets unless counter‑balanced by regulation or data portability increasing worker bargaining power.

• Productivity, effort and monitoring

  • Human‑tech systems (computer vision, route/warehouse optimization) raise measurable productivity but can harden targets and transfer risk to workers (weather, traffic). Empirical models should separate technical efficiency gains from distributional effects (who captures rents).
  • Costs of implementing human‑in‑the‑loop (HITL) controls — e.g., human review of deactivations — raise operational costs for platforms and create tradeoffs between speed/coverage of automation and fairness; these costs will affect platform pricing and labor market equilibrium.

• Data as an economic asset and worker bargaining

  • Proposals like the Unified Worker Interface (UWI) would reduce platforms' exclusive data rents by enabling portable verified worker data and reputations, potentially increasing worker marketability and changing equilibrium wages and turnover.
  • Creating a Data Gigs ecosystem could resegment labor supply toward digital microtask markets, altering cross‑sector labor elasticities and creating complementarities between location‑based gigs and remote digital tasks.

• Policy interventions as economic shocks / natural experiments

  • State‑level gig laws (transparency, explainability, grievance norms) provide staggered policy variation ideal for causal identification. Economists can estimate effects on earnings volatility, deactivation rates, platform labor supply, and consumer prices using differences‑in‑differences and regression discontinuity.
  • Enforcement of human review and right to explanation changes platform enforcement behavior; models should capture strategic platform responses (e.g., stricter upfront screening versus more human resources for review).

• Measurement and empirical priorities

  • Need for granular platform administrative data (orders, prices, incentives, penalties, deactivation logs) linked to worker outcomes to quantify income volatility, spell lengths after deactivation, and marginal returns to time.
  • Field experiments and audit studies to test for algorithmic bias (allocation by geography/demographics), incentive manipulation, and transparency interventions (e.g., showing explicit surge formula) on worker behavior.
  • Survey and panel data to measure worker understanding of algorithmic rules, risk aversion, and labor supply elasticity under different disclosure regimes.

• Broader welfare and distributional effects

  • Algorithmic management can widen inequality when gains accrue to capital/tech owners while exposing workers to higher variance pay and opaque sanctions. Evaluations should quantify welfare changes under counterfactual regulatory or data‑infrastructure scenarios (e.g., UWI rollout, mandated human review).

In sum, this study highlights that algorithmic systems shape not just productivity but important economic frictions (information asymmetry, income risk, enforcement externalities) in platform labor markets. AI‑economics research should model these mechanisms explicitly, exploit policy heterogeneity for causal inference, and evaluate welfare tradeoffs of transparency, human‑in‑the‑loop requirements, and data portability interventions.