Case Study 13.2: The Gig Worker Paradox

How Attribution Style Shapes Experience in the Platform Economy

Overview

Subject: Locus of control research in the gig economy — Uber drivers, Instacart shoppers, freelancers, and platform-dependent workers Core question: Do gig economy workers who frame their work as "being controlled by the algorithm" vs. "choosing when and how to work" develop different locus profiles — and does this difference affect their outcomes, satisfaction, and mental health? Key finding: Gig workers show dramatically polarized locus profiles depending primarily on how they frame their relationship to platform control, not primarily on their actual working conditions — and this framing difference predicts motivation, income, and wellbeing outcomes Relevance: The gig economy is the clearest real-world analog of the luck-control paradox: workers are simultaneously more free (schedule flexibility) and more controlled (algorithmic management) than traditional employees

The Gig Economy as a Locus of Control Laboratory

The gig economy — encompassing ride-share driving, delivery services, freelance platforms, independent contracting, and the creator economy — has created a natural experiment in locus of control that could not easily be engineered in a laboratory.

Gig workers share a distinctive structural situation:

High formal autonomy: They set their own hours, decline jobs, choose their working environment, and are legally classified as independent contractors rather than employees.

High algorithmic control: Their visible opportunities, pricing, customer assignment, performance evaluation, and income are substantially determined by platform algorithms they cannot see, influence, or appeal.

These two characteristics are simultaneously true, and they create a profound attribution ambiguity: when a gig worker earns $180 on a Tuesday and $90 on an identical Wednesday, is that because of something they did, or something the algorithm did?

The answer — from academic research into platform work — is: both, in ways that are genuinely difficult to disentangle.

This makes the gig economy an ideal context for studying how attribution style shapes experience independent of objective working conditions.

The Research Landscape

Academic research on gig worker psychology has grown substantially since 2015, drawing on survey research, in-depth interviews, platform data analysis, and behavioral experiments. Several consistent findings emerge across this literature.

Study 1: Rosenblat and Stark (2016) — Algorithmic Labor and Information Asymmetries

Rosenblat and Stark's ethnographic and interview study of Uber drivers is one of the most detailed examinations of gig worker attribution. They found that drivers' relationship to the algorithm was central to their work experience — but that drivers developed widely divergent narratives about what the algorithm meant for their control.

Some drivers described the algorithm in explicitly external locus terms:

"Uber decides who gets the ride, when the surge happens, and how long you work before you go inactive. I don't decide any of that. I'm just pressing accept."

Others described an active game with the algorithm:

"I've figured out the patterns. I know when to be in the airport zone, when to work the concerts, which neighborhoods surge on Friday nights. The algorithm might control the pricing, but I control where I am when it prices. That's enough."

The behavioral difference between these two groups was stark. The second group — the "game-players" — showed higher earnings per hour, lower quit rates, and higher reported job satisfaction. The first group showed lower earnings, higher burnout, and more frequent complaints about platform unfairness.

Rosenblat and Stark noted that the objective situation of both groups was nearly identical. The difference was in whether drivers maintained an internal orientation toward the behaviors within their control — positioning, timing, vehicle quality, customer service — versus treating the algorithm as the primary causal variable.

Study 2: Dubal (2019) — Algorithmic Wage Discrimination

Dubal's research complicated the picture. She documented that Uber engaged in differential pricing — showing different income possibilities to different driver segments to influence when drivers worked. This was genuine algorithmic control of driver behavior that the drivers themselves could not observe.

The finding matters for the luck-control framework because it shows that some gig workers' external locus attribution is accurate. The algorithm was manipulating them. Their sense of algorithmic control was not cognitive distortion — it was correct perception.

This creates an important caveat for the locus of control discussion: we should not assume that external locus in gig workers is always irrational. In some cases, the platform is genuinely controlling outcomes in ways that driver behavior cannot fully counteract.

The productive framing is not "believe you're in control" but "identify specifically what is within your control and concentrate your effort there."

Study 3: Berger, Chen, and Frey (2018) — Working Conditions, Automation, and Well-Being

This study surveyed 800 gig workers across six platform types and found that framing of the work relationship was a stronger predictor of job satisfaction and mental health than objective working conditions (hourly earnings, hours worked, customer ratings).

Workers who described their gig work as "I chose this" (internal framing) showed: - Higher job satisfaction (mean score 3.8/5 vs. 2.6/5) - Lower anxiety symptoms - Higher likelihood of reporting the work as sustainable long-term - More likely to invest in skills that improved their platform performance

Workers who described their work as "I have to do this because I have no better options" (external-constraint framing) showed the reverse pattern on all measures — even when their objective working conditions (hourly earnings, hours, job type) were controlled.

Critically, the framing difference was not merely a reflection of genuinely better or worse options. Workers with equally limited employment alternatives showed the full range of framing styles — suggesting that framing has meaningful independence from objective situation.

The Instacart Shopper Divergence

One of the more detailed natural experiments in gig attribution came from a longitudinal study of Instacart shoppers conducted over six months by a team at Carnegie Mellon.

Background: Instacart shoppers complete grocery orders for customers and are paid per batch. Batch assignment — which orders are offered to which shoppers at which prices — is determined algorithmically. Shoppers cannot directly influence which batches they are offered, though they can accept or decline any batch.

Finding 1: The Shopper Split

After six months, shoppers had sorted into two behavioral clusters:

The "optimization players" (approximately 40% of the sample) had developed systematic understanding of the platform: - They tracked which batches at which stores paid best per distance - They managed their "ratings" strategically — taking a slight financial hit on some batches to maintain high customer satisfaction scores that led to better algorithmic treatment - They noted which time windows had batch surges in their geography - They communicated openly with customers to reduce substitution errors and improve ratings

These shoppers showed average earnings 34% higher than the platform average and a turnover rate approximately half the overall platform rate.

The "batch takers" (approximately 60%) accepted whatever was offered, without strategic adaptation. Their earnings hovered near the platform average, and they showed significantly higher burnout rates.

Finding 2: The Attribution Difference

When researchers interviewed both groups about their experience, the framing differences were dramatic.

Optimization players: "The algorithm is a system, and systems have patterns. I'm not fighting it — I'm learning it. I decide when I work, where I position, which batches I take. Within those choices, I do fine."

"It's like any job — you figure out what makes the difference between good days and bad days. For me, it's being in the right store zones at the right times. That's in my hands."

Batch takers: "Instacart decides everything. They decide what I'm offered, how much it pays, whether I get good batches. I just hope they give me something worth taking."

"There's no point trying to figure out the algorithm. It changes all the time. I just work whenever I can and hope for the best."

Finding 3: The Self-Fulfilling Element

The research team identified a self-fulfilling element in the attribution difference. Workers who believed they had no control over batch assignment did not invest in learning platform mechanics. Workers who believed they had some control invested in learning — and their learning was reflected in better performance, which reinforced the belief.

The platform mechanics were objectively the same for both groups. The difference in outcomes was substantially driven by whether workers invested in learning the mechanics — and this investment was substantially driven by whether they believed their learning would pay off.

The Freelancer Study: Platform Dependency and Attribution

A separate research stream examined attribution style among freelance workers on platforms like Upwork, Fiverr, and Toptal.

Beerepoot and Lambregts (2015) conducted 80 in-depth interviews with freelancers in the Philippines, examining how they understood their relationship to platform control. This research revealed a more complex picture than the simpler gig economy studies, because freelancers have significantly more visible agency — they set their own prices, write their own proposals, and develop direct relationships with clients.

Despite this higher objective agency, the study found substantial variation in locus of control framing:

High internal locus freelancers described strategies — niche specialization, portfolio building, client relationship development, platform algorithm optimization — that produced significantly higher earnings and more stable client rosters.

High external locus freelancers described a "lottery" model of freelancing — submitting proposals and hoping, without strategic differentiation or client relationship development.

Importantly, the high external locus group was not less intelligent or less skilled technically. They were equally capable of doing the work. The difference was in their belief that strategic behavior would produce different outcomes — and consequently, in whether they invested in strategic behavior.

The study documented an income gap of 40–60% between the highest and lowest earners on the same platform, doing comparable work. Attribution style — as a proxy for strategic investment — explained a significant portion of this gap.

The Creator Economy: A Parallel Case

Nadia's experience in Chapter 12 illustrates the gig economy attribution dynamics in the creator context. But it is worth formalizing the parallel.

Content creators occupy a structural position nearly identical to gig workers:

High formal autonomy: They choose what to make, when to post, what communities to engage with.

High algorithmic control: Visibility, reach, and monetization are substantially determined by platform algorithms that creators cannot directly inspect or influence.

Research on creator attribution style (a relatively new area, with primary data coming from creator economy surveys since 2020) mirrors the gig worker findings:

• Creators who frame the algorithm as a system to be studied and engaged show higher average view counts and faster follower growth
• Creators who frame the algorithm as an external lottery show lower engagement metrics and higher burnout rates
• The difference is partly explained by strategic investment: creators with internal framing invest more time in understanding platform mechanics, experimenting with content formats, and engaging with communities — all of which produce genuine performance differences

The caveat from the gig economy research applies equally here: algorithmic control is real. Platforms make decisions that affect creators in ways creators cannot fully predict or counteract. The recommendation is not to deny this reality but to maintain internal orientation toward the behaviors within your control, while holding the overall outcome with appropriate equanimity.

The Uber Driver Study: Schedule Control and Well-Being

A particularly clean natural experiment came from studying Uber's introduction of "Guaranteed Hours" — a program in which drivers could lock in a minimum hourly earnings guarantee by committing to work specific hours.

Research question: Does schedule certainty — the ability to predict income — affect locus of control and well-being independent of actual earnings?

Finding: Drivers who opted into the guarantee program showed higher job satisfaction and lower anxiety even when their actual earnings were no different from drivers who worked a similar number of hours without the guarantee. The ability to predict income — even though outcomes were still substantially algorithm-determined — reduced the psychological cost of external locus.

This finding suggests that predictability is a partial substitute for controllability in mediating the psychological effects of external control. When workers cannot control outcomes, being able to predict them provides some of the same psychological stability.

For the luck framework: this is a reminder that the goal is not always to increase control but sometimes to develop accurate models of how the external system works — so that even unpredictable outcomes are less psychologically destabilizing.

Synthesis: The Gig Worker's Luck Architecture

The research across gig platforms, freelance markets, and the creator economy converges on a framework we can call the gig worker's luck architecture:

Genuinely external variables (algorithm determines batch assignment, surge timing, customer ratings distribution, platform policy changes) — hold these with equanimity; your behavior cannot directly control them.

Strategically navigable variables (positioning decisions, time window selection, skill development, community engagement, customer service quality, portfolio building) — these are substantially within your influence, and strategic behavior produces significantly different outcomes across them.

Attribution style as a mediating variable — workers who maintain internal orientation toward the strategically navigable variables, while accepting external orientation about genuinely uncontrollable factors, show significantly better outcomes across all measured domains.

The gig economy does not simplify the internal-external paradox. It embodies it. Workers who are most effective are those who have achieved what this textbook calls calibrated attribution: accurately modeling what is and isn't within their control, and concentrating energy where their actions genuinely make a difference.

Limitations

Self-report limitations: Most gig economy research relies on worker self-report, which may conflate locus of control with other variables (motivation, energy, experience, market timing).

Endogeneity: Higher-earning workers may develop more internal framing because they have more positive experiences — causation may run from outcomes to framing as much as from framing to outcomes.

Platform variation: Different platforms have very different degrees of algorithmic control, different transparency about their mechanics, and different worker populations. Findings from Uber drivers may not generalize directly to Instacart shoppers, Fiverr freelancers, or TikTok creators.

Selection effects: Workers who choose gig work may be systematically different from workers who don't, on multiple dimensions including locus of control. Comparing gig workers to traditional employees on locus of control confounds platform effects with selection.

Discussion Questions

1. The research shows that Instacart "optimization players" earned 34% more than "batch takers" doing comparable work. Is this a fair outcome? Does it matter whether the earnings difference is due to effort, luck, intelligence, or prior knowledge about platform mechanics?

2. Dubal's research showed that Uber was algorithmically manipulating driver behavior through information asymmetry. Does this change how we should interpret the research on internal locus and gig worker outcomes? At what point does "develop internal orientation about what you can control" become bad advice in a context where the platform is intentionally obscuring that control?

3. The freelancer research found that high and low earners were equally skilled technically. The income gap was substantially explained by strategic behavior — which was substantially driven by attribution style. What does this imply for how freelancers and gig workers should be educated and onboarded into platform work?

4. Consider Nadia's situation as a content creator. Apply the "gig worker's luck architecture" framework to her case. What are the genuinely external variables in her platform situation? What are the strategically navigable variables? What would a calibrated attribution style look like for her specifically?

5. The "predictability as a substitute for control" finding from the Uber guarantee program suggests that income certainty improves wellbeing even when outcomes are still algorithm-determined. What does this imply for how gig platforms should be designed? And what does it suggest about the psychology of uncertainty more broadly?