AI and the Gig Economy

The gig economy runs on a simple premise: connect workers with customers, take a cut, avoid being an employer. But this model only works at scale because of AI. Without algorithmic systems to allocate work, set prices, evaluate performance, and discipline workers, platforms like Uber, DoorDash, and Instacart couldn't function.

What emerges is a new form of employment where the app is the manager. The workers remain — driving, delivering, shopping, cleaning — but the traditional workplace relation has been stripped out and replaced with software. This isn't automation of work. It's automation of the employment relationship itself. The pattern echoes centuries of technological mediation of labor: from factory clocks to assembly lines to electronic monitoring, each wave redefines the boundary between worker autonomy and employer control. What's different now is the speed and granularity of that control.

The Platform Model: AI as Manager, Intermediary, and Market-Maker

Platform companies position themselves as technology providers, not employers. They build apps that connect supply (workers) with demand (customers). The platform facilitates; it doesn't employ.

This framing serves a legal purpose: avoiding employment law obligations. But it obscures what platforms actually do. They don't merely connect parties — they manage labor through algorithmic systems that perform core employer functions.

Three functions platforms perform through AI:

Manager: The algorithm assigns work, evaluates performance, sets expectations, and metes out consequences. It decides which driver gets which ride, which shopper gets which order, which cleaner gets which job. This is management — it's just been encoded in software rather than delegated to human supervisors.

Intermediary: The platform sits between worker and customer, controlling information flow in both directions. Workers can't see customers directly; customers can't negotiate with workers directly. All communication routes through the platform, which extracts data from every interaction.

Market-maker: The algorithm sets prices, matching supply and demand in real time. Surge pricing, dynamic rates, time-sensitive bonuses — these mechanisms determine what workers earn. The platform creates the market, then controls its parameters.

Steven Vallas and Juliet Schor, in their 2020 analysis "What Do Platforms Do?", identify this as a distinctive form of economic organization. Platforms aren't simply firms or markets — they're something new, combining elements of both while claiming the privileges of neither.

Information Asymmetries: What Platforms Know vs. What Workers Know

The defining feature of gig platform AI is information asymmetry. Platforms see everything; workers see almost nothing.

Alex Rosenblat's Uberland (2018) documents this asymmetry in detail. When an Uber driver receives a ride request, they see limited information: pickup location, rider rating, perhaps an estimate of trip direction. They don't see the destination, trip distance, expected fare, or why this particular ride was sent to them rather than another driver.

The platform, meanwhile, has complete visibility:

• Driver location, speed, and driving patterns
• Historical acceptance and completion rates
• Ratings from every past ride
• Device information, app activity, breaks taken
• Passenger information, trip patterns, payment data
• Market-wide supply and demand in real time

This asymmetry isn't incidental — it's designed. If drivers could see all available rides, they'd cherry-pick the profitable ones. If they understood the matching algorithm, they'd optimize against it. The platform's power depends on workers navigating blind while the system sees everything.

Rosenblat identifies specific mechanisms of information control:

Destination concealment: Uber drivers don't know where a ride will take them until they accept. This prevents rational economic decisions — a driver might accept a ride that takes them to a low-demand area, stranding them without return fares.

Acceptance rate pressure: Drivers who decline rides face consequences: lower priority for future assignments, warnings, eventual deactivation. But declining might be rational — a pickup might be far away, a destination might be unsafe. The system punishes prudent decision-making.

Rating opacity: Drivers know their overall rating but not which rides produced which scores or why. A passenger might give one star for reasons unrelated to driver performance. The driver can't know, can't contest, can't learn.

Algorithmic suggestions: The app "suggests" when to drive, where to position, which bonuses to pursue. These suggestions carry implicit coercion — ignore them and your earnings suffer.

The result: workers make consequential decisions (when to work, which jobs to accept, how to perform) without the information needed to make those decisions well. The platform captures informational advantage as a source of control.

Algorithmic Pricing: Surge, Dynamic Rates, and Opaque Pay

How much does a gig worker earn? The honest answer: it depends, and the determinants are opaque.

Veena Dubal's 2023 paper "Algorithmic Wage Discrimination" argues that gig platforms have invented a new form of wage-setting. Traditional employers offer wages — hourly rates, salaries, piece rates. These may be low or high, fair or exploitative, but they're knowable in advance. Workers can compare offers and make informed choices.

Platform pricing works differently. The same task — the same ride, the same delivery — might pay vastly different amounts depending on:

• Time of day and day of week
• Local supply and demand conditions
• The individual worker's history and predicted behavior
• Promotional campaigns and incentive structures
• Factors the platform doesn't disclose

Dubal documents cases where two Uber drivers completing identical trips at the same time received different payments. The platform's black-box pricing makes such discrimination invisible and unchallengeable.

Surge pricing:

When demand exceeds supply, prices rise — and worker pay rises with them. In theory, this allocates workers efficiently: high prices attract drivers to busy areas.

In practice, surge creates perverse dynamics:

• Workers learn to manipulate surge by going offline collectively, triggering artificial scarcity
• Platforms may induce phantom surges to manipulate worker behavior
• Passengers bear the cost of supply shortages that platforms could address through guaranteed minimums

Dubal notes that surge pricing transfers risk from platforms to workers and customers. The platform guarantees nothing; workers and customers absorb variability.

Upfront pricing:

Some platforms now show workers expected pay before they accept a job. This seems like transparency — but the algorithm determining that pay remains opaque. Workers know what they'll earn on this job but not why that number was chosen or what alternatives they forewent.

DoorDash's "Dasher Pay Model" exemplifies the complexity. Base pay varies by zone, distance, and "desirability" — an undefined metric. Tips are sometimes included in upfront quotes, sometimes separate. Promotional bonuses appear and disappear. Workers report that effective hourly rates fluctuate wildly, and understanding the algorithm's logic proves impossible.

The redistribution lens:

Traditional employment involves negotiation — individual or collective — over compensation. Platforms have replaced negotiation with algorithmic dictation. Workers can accept the platform's terms or exit; there is no bargaining.

This transfers power from workers to platforms. The efficiency gains from dynamic pricing accrue primarily to platform operators, not workers. When the algorithm gets it right, the platform captures the surplus. When it gets it wrong, workers bear the loss. As Daron Acemoglu and Simon Johnson argue in Power and Progress, this is not an inevitable feature of technology but a consequence of who controls its design.

Rating Systems and Reputational Control

Your rating is your livelihood. Gig workers navigate a world where a number — 4.8, 4.6, 4.2 — determines economic survival.

Platform rating systems serve multiple functions:

Quality signal: Ratings ostensibly tell customers which workers are reliable. A five-star driver is probably better than a four-star driver.

Management without managers: Ratings discipline workers without human supervisors. Low ratings trigger warnings, reduced assignments, eventual termination. The algorithm enforces standards through aggregated customer feedback.

Control mechanism: The threat of bad ratings shapes worker behavior. Drivers tolerate abusive passengers because pushing back risks a rating hit. Workers perform emotional labor — friendliness, deference, patience — that wouldn't be required in a traditional job.

Problems with the rating model:

The five-star scale creates brutal dynamics. On most platforms, anything below 4.6 is concerning; below 4.3 risks deactivation. The functional range isn't 1-5 but 4.3-5.0 — a narrow band where small fluctuations determine economic fate.

Ratings reflect factors beyond worker control:

• Traffic that delayed a delivery
• Restaurant slowness that made food cold
• Customer mistakes in ordering
• Prejudice against workers' race, gender, accent, appearance

Research documents rating bias. Studies show that women and people of color receive systematically lower ratings for equivalent service. The "objective" rating system encodes subjective discrimination.

Workers can't contest individual ratings or understand their sources. A driver doesn't know which ride produced which score. This makes improvement impossible — you can't fix what you can't identify.

Reputational lock-in:

Ratings don't transfer between platforms. A five-star Uber driver starts at zero on Lyft. This creates switching costs that bind workers to platforms — one form of control substituting for employment contracts.

Rosenblat describes this as "platform-specific human capital." Workers invest in building reputations that only that platform values. The platform captures the benefit; the worker bears the investment.

Worker Classification: The Core Legal Battle

Are gig workers employees or independent contractors? This question — fought in courts, legislatures, and ballot initiatives — determines access to minimum wage, overtime, workers' compensation, unemployment insurance, and collective bargaining rights.

The platform position:

Platforms argue workers are independent contractors because:

• They set their own hours
• They can work for multiple platforms
• They use their own vehicles/equipment
• They're free to accept or decline work

This framing emphasizes flexibility and autonomy. Workers aren't controlled; they're empowered.

The counter-argument:

Critics note that platforms exercise control through algorithmic mechanisms:

• They set prices workers can't negotiate
• They monitor performance continuously
• They impose consequences for ratings and acceptance rates
• They control access to work through opaque matching
• They can deactivate accounts unilaterally

The flexibility is real but constrained. Workers can choose when to work, but they can't choose how work is compensated or allocated. The significant decisions rest with the platform.

California's AB5 and Proposition 22:

California's AB5 (2019) codified a stricter test for contractor status: workers are employees unless the hiring entity proves they're free from control, perform work outside the usual course of business, and have an independent business of their own.

Gig platforms failed this test. Uber drivers drive — driving is Uber's core business.

In response, platforms spent over $200 million on Proposition 22 (2020), a ballot measure exempting app-based drivers and couriers. It passed with 59% of the vote, creating a hybrid category with some benefits but no employment status.

The result: California gig workers receive healthcare stipends and accident insurance but can't unionize, lack unemployment benefits, and have limited recourse against termination. The proposition also requires 7/8 of the legislature to amend it — an extraordinarily high bar.

EU Platform Work Directive (2024):

The EU took a different approach. The directive creates a presumption of employment when platforms meet certain criteria indicating control. Workers aren't contractors by default; employers must prove contractor status is appropriate.

Member states are still implementing the directive, and platforms are restructuring to evade classification triggers. But the legal burden has shifted.

UK and Spanish courts:

The UK Supreme Court ruled in 2021 that Uber drivers were "workers" (a UK-specific category between contractor and employee) entitled to minimum wage and holiday pay. Spain's "Rider Law" (2021) presumed delivery riders are employees.

These cases demonstrate that platform classification isn't technically determined — it's politically contested, resolved differently across jurisdictions.

Surveillance: GPS, Facial Recognition, and Activity Monitoring

Gig platforms are surveillance companies. Their business model depends on watching workers continuously.

Location tracking:

Every gig platform tracks worker location via GPS. This enables core functionality — you can't dispatch a driver without knowing where they are. But location data serves control purposes too:

• Platforms monitor speed, routes, and stop duration
• Deviations trigger algorithmic flags
• Workers can be penalized for "inefficient" routes even when avoiding traffic or hazards
• Break times are monitored — extended stops may count against metrics

Drivers report the anxiety of constant observation. You can't stop for a bathroom break without considering how it affects your numbers.

Facial recognition:

Some platforms use facial recognition to verify worker identity. Uber's "Real-Time ID Check" requires drivers to take selfies that match their account photo. The stated purpose: preventing fraud and ensuring the person driving is the person authorized.

But facial recognition creates problems:

• Recognition systems show higher error rates for darker-skinned faces
• Workers report being locked out of accounts due to mismatches
• The requirement to submit biometric data raises privacy concerns
• There's no transparency about where this data goes

Algorithmic surveillance without cameras:

Even without cameras, platforms monitor behavior through patterns:

• How long between accepting a job and starting it?
• How does speed compare to expected travel time?
• How often do workers cancel after accepting?
• What does the worker's phone usage indicate about attention?

This data informs algorithmic decisions — assignment priority, rating algorithms, deactivation triggers — without workers knowing what's measured or how.

The panopticon effect:

Workers must assume they're being watched, even when they can't see the watcher. This shapes behavior beyond explicit rules. The surveillance is productive — it creates discipline without direct management.

Michel Foucault's concept of the panopticon applies precisely. The guard tower in the prison's center might be empty, but prisoners behave as if watched because they can't tell. Platform workers live this logic daily.

Resistance: Organizing, Litigation, and Regulation

Workers haven't accepted algorithmic control passively. Resistance has taken multiple forms.

Informal resistance:

Workers share information about platform algorithms through forums, Facebook groups, and in-person conversations. They trade tips about when surge happens, which jobs to avoid, how to game the rating system. This collective intelligence pushes back against informational asymmetry.

Some workers coordinate collective action without formal organization. Driver "log-offs" — coordinated moments when drivers go offline simultaneously — can trigger surge pricing or protest platform changes. These flash actions leverage the platform's dependence on labor supply.

Formal organizing:

Traditional union organizing is difficult in gig work. Workers are geographically dispersed, have no common workplace, and are classified as contractors without labor law protections.

But organizing happens anyway:

• Independent Workers' Union of Great Britain (IWGB): Represents Uber drivers, Deliveroo riders, and other gig workers. Won the 2021 UK Supreme Court case establishing worker status.
• Los Deliveristas Unidos (New York): Delivery workers organized for minimum pay protections passed in 2023.
• Gig Workers Rising (California): Coalition organizing across platforms, advocating for AB5 and against Proposition 22.
• Riders X Derechos (Spain): Riders' collective action led to the 2021 "Rider Law" establishing employment status.

These efforts face structural obstacles. Platform workers can't picket a workplace that doesn't exist. Turnover is high; building durable membership is hard. And platforms can respond to organizing by changing algorithms in ways that penalize activists.

Litigation:

Legal challenges have produced significant wins:

• Aslam v. Uber (UK Supreme Court, 2021): Uber drivers are workers entitled to minimum wage and holiday pay
• Dynamex (California Supreme Court, 2018): Established the ABC test later codified in AB5
• Delivery rider cases in Spain, Italy, France: Various courts found employment relationships

But litigation is slow, expensive, and vulnerable to platform adaptation. Uber responded to the UK ruling by tweaking its model and disputing that it applies to current operations. Legal wins don't automatically translate to changed conditions.

Regulatory approaches:

Beyond classification battles, some jurisdictions have enacted targeted regulations:

• New York City minimum pay: Delivery workers must receive minimum per-trip and per-minute rates.
• Seattle: Gig workers gained access to the city's sick leave ordinance.
• EU Platform Work Directive: Requires transparency about algorithmic decision-making, human oversight of significant decisions, and data access for worker representatives.

The regulatory landscape remains fragmented. What protects workers in Barcelona doesn't apply in Dallas. Platforms operate globally while labor law remains local.

What Gig Work Reveals About AI and Labor

The gig economy isn't a peculiar corner of the labor market. It's a laboratory for practices that may spread much further.

The disaggregation model:

Platforms proved you can separate tasks from employment. Work gets done, but there's no employer-employee relationship in the traditional sense. As AI enables more fine-grained task allocation and monitoring, this model could extend to occupations currently structured as jobs.

Professional services, white-collar work, even management itself might be "gig-ified" — tasks allocated by algorithm, performance evaluated automatically, workers engaged without employment. The gig economy pioneered the template. The dynamics already visible in White-Collar Displacement and Freelance Markets After Generative AI follow the same logic, applied to different skill levels.

Algorithmic management goes mainstream:

The techniques developed for gig work are migrating to traditional employment. Amazon warehouses use algorithmic monitoring, automated warnings, and machine-speed performance expectations. Call centers score calls with AI and coach workers in real time. Even office work increasingly involves AI surveillance of email, calendar, and productivity metrics.

The gig economy proved algorithmic management works — meaning it extracts labor effectively — even if workers don't like it. That proof of concept invites broader adoption. Though it is worth noting that some employers are exploring more transparent models: companies like Alto and some regional cooperatives demonstrate that ride-hailing can operate with employee drivers and algorithmic transparency, even if they haven't yet matched platform scale.

The information asymmetry model:

Gig platforms demonstrated that controlling information is a source of power. Keep workers ignorant of the algorithm, and they can't organize against it, optimize around it, or negotiate over it. This model applies far beyond gig work — any employer deploying AI can choose transparency or opacity.

The risk transfer model:

Gig work transfers risks from platforms to workers: demand variability, vehicle maintenance, insurance, healthcare, retirement. When work is slow, the platform bears no cost; the worker simply earns nothing. This risk redistribution has proven lucrative for platforms and may tempt other employers to shift toward contingent arrangements.

The Redistribution Question

Gig platforms redistribute value in several directions — mostly away from workers.

Labor-to-capital:

Workers perform the productive activity — driving, delivering, shopping, cleaning. The platform provides software infrastructure and matching. But the platform captures most of the surplus through fees that typically range from 15-30% of transaction value. Workers bear the costs (vehicles, gas, insurance, time) while platforms extract recurring cuts.

Risk-to-workers:

Traditional employment bundles work with protections: guaranteed wages, benefits, job security. Gig work unbundles them. Workers get tasks; platforms keep flexibility. When demand dips, platforms aren't paying idle workers. When workers get injured or sick, platforms owe nothing. The risks of economic life transfer from business entities that could absorb them to individuals who can't.

Information-to-platforms:

Every gig interaction generates data: location, timing, performance, customer behavior. This data improves platform algorithms and informs business decisions. Workers generate it; platforms own it. There's no compensation for the data value workers create through their labor.

Accountability-to-nowhere:

When something goes wrong — discrimination, safety incidents, unfair terminations — accountability is diffuse. The algorithm made the decision. The platform didn't intend the outcome. Individual responsibility dissolves into system design, and the system's designers are shielded from consequences.

What redistribution in the other direction would look like:

• Minimum pay guarantees: Workers earn at least a floor regardless of algorithmic whims.
• Algorithmic transparency: Workers understand and can contest the systems governing them.
• Data rights: Workers own or share the data their labor generates.
• Portable benefits: Healthcare, retirement, workers' comp that follow the worker, not the employer.
• Collective bargaining: Sectoral standards negotiated between workers and all platforms in an industry.
• Employment status: For workers who function like employees, the legal protections of employment.

Some of these reforms exist in some jurisdictions. None exist universally. The gig economy's current structure reflects a political victory by platforms — one that could be reversed through countervailing power. The collective bargaining victories in other sectors demonstrate that workers can win meaningful protections when organized. And the cooperative platform movement — from driver-owned apps like The Drivers Cooperative in New York to platform cooperativism as theorized by Trebor Scholz — offers structural alternatives where workers share governance and surplus. These aren't utopian; they're operating businesses that prove the gig model is a choice, not an inevitability.

What Should Readers Know?

If you use gig platforms: The price you pay includes what workers give up — employment protections, stable income, recourse against unfair treatment. The convenience is real, but so are the costs, and the costs fall on someone else.

If you work in the gig economy: The algorithmic systems governing you aren't neutral or inevitable. They're designed choices that serve platform interests. Connecting with other workers — through informal networks or formal organizations — is the most effective counter-pressure.

If you shape policy: Worker classification is necessary but not sufficient. Even reclassified workers need transparency rights, algorithmic accountability, and collective bargaining power to negotiate effectively with systems this complex.

If you build AI systems: Gig platforms are a preview of what happens when AI governs labor without accountability. The technical architecture matters — opacity vs. transparency, unilateral control vs. worker voice. These aren't neutral engineering choices; they're political decisions embedded in code.

Related

• Manager Is Algorithm
• Hidden Labor
• Automation Debate
• Collective Bargaining
• White-Collar Displacement
• Freelance Markets
• AI Cooperatives
• Power and Progress
• Automating Inequality
• EU AI Act Takes Effect

Sources

• Dubal, Veena. "Algorithmic Wage Discrimination." Columbia Law Review 123, no. 7 (2023). Columbia
• Rosenblat, Alex. Uberland: How Algorithms Are Rewriting the Rules of Work. University of California Press, 2018. Publisher
• Vallas, Steven P. and Juliet B. Schor. "What Do Platforms Do? Understanding the Gig Economy." Annual Review of Sociology 46 (2020). Annual Reviews
• Mateescu, Alexandra and Aiha Nguyen. "Algorithmic Management in the Workplace." Data & Society, 2019. Data & Society
• Gray, Mary L. and Siddharth Suri. Ghost Work: How to Stop Silicon Valley from Building a New Global Underclass. Houghton Mifflin Harcourt, 2019. Publisher
• Uber BV v. Aslam [2021] UKSC 5. UK Supreme Court
• European Parliament. "Platform Work Directive." 2024. EUR-Lex
• State of California. "Assembly Bill 5 (AB5)." 2019. California Legislature
• State of California. "Proposition 22." 2020. Official Voter Information Guide
• Government of Spain. "Ley Rider (Royal Decree-Law 9/2021)." 2021. BOE
• De Stefano, Valerio. "The Rise of the Just-in-Time Workforce." Comparative Labor Law & Policy Journal 37, no. 3 (2016). SSRN
• Kellogg, Katherine C., Melissa A. Valentine, and Angèle Christin. "Algorithms at Work: The New Contested Terrain of Control." Academy of Management Annals 14, no. 1 (2020). AoM