Case Study 1.6: Robo Firing

4.4 Million

Gig economy workers in the UK

10,169

London drivers robo-fired in 1 year (+123%)

94% BAME

London private hire drivers facing biased AI

230 Rows/Min

Device data tracked by Uber even offline

1 Case Study & Background: What Social Problem is Being Solved?

What is our case study about? Our case study examines the Worker Info Exchange (WIE) report "Managed by Bots: Data-Driven Exploitation in the Gig Economy." It reveals how platforms like Uber, Deliveroo, Ola, Bolt, Free Now, Just Eat, and Amazon Flex use automated bots to manage and unfairly dismiss workers—a practice known as "Robo Firing."
Economic Context & Unpaid Time: For years, platforms misclassified workers as "independent contractors" to avoid paying minimum wage, holiday pay, or taxes. A New York study showed drivers spend only 58% of their working time on paid passenger trips; the remaining 42% is unpaid waiting time on the app.
Misclassification 2.0 (Hiding Control Behind Algorithms): After the UK Supreme Court ruled that Uber drivers are actually workers under Uber's strict control, platforms adopted "Misclassification 2.0." Instead of using human managers, they hide strict management control behind automated algorithms and bots so workers cannot claim full labor rights.

🏢 The Platform View

Deliveroo states in its privacy policy that checking millions of deliveries manually is impossible for humans, so automated bots are needed to stop fraud and contract violations. Similarly, Uber rushed facial recognition and GPS checks in London just to stop 21 account-sharing cases out of 90,000 drivers.

✊ The Worker & WIE View

The real social problem is algorithmic exploitation and mass Robo Firing without fair human hearings (affecting 20% of active London drivers in one year). WIE solves this by exercising GDPR data rights, establishing a "Worker Data Trust", and using strategic litigation to build collective bargaining power.

2 Relevant Course Themes & Questions Posed

Datafication of Work & Algorithmic Management: Directly connected to our course leader Gwendolin Barnard’s research, platforms convert every worker's physical labor into three distinct layers of data:

1. Input Data

Information provided directly by the worker (ID cards, driving licenses, vehicle documents).

2. Observed Data

Raw data captured while using the app (GPS location, telematics, vehicle speed, braking style).

3. Inferred Data

Secret AI-generated profile scores (e.g., "Fraud probability scores" and "Earning profiles").

Technology is Political (Langdon Winner, 1986): As discussed in our first lecture, technology is never neutral; power and politics are embedded directly inside algorithmic code. Robo firing is not an accidental technical glitch—it is a political weapon designed to keep workers subordinate and bypass labor laws.

3 Inequality and Discrimination

⚠️ Surveillance Case Study I: Racial Bias in AI (Pa Edrissa Manjang)

Just like Slide 17 in our lecture (Tarcizio Silva's tweet showing computer vision labeling an object as a "gun" in a Black hand and a "tool" in a white hand), Uber’s system exhibits severe racial bias. In 2020, Uber deployed Microsoft’s FACE API for Real-Time ID (selfie) checks. Although 94% of London private hire drivers are Black and ethnic minority (BAME), this technology has high error rates on darker skin tones. Pa Edrissa Manjang, a Black Uber driver, was robo-fired for a "failed selfie match" and denied human review. When WIE recovered his photos via a DSAR, every single selfie was clearly his! Even Microsoft admits the AI makes errors and requires human review, which Uber ignored.

Informational Asymmetry: Platforms hoard all workplace data while blocking workers from seeing it, creating a massive power imbalance.
Discriminatory Work Allocation & Dynamic Pricing: A court in Bologna, Italy ruled Deliveroo's rating algorithm discriminatory because it penalized workers for illness or legal strikes. Furthermore, Uber and Ola use past behavior for "Upfront Dynamic Pricing"—creating the risk that algorithms identify desperate, low-income drivers and intentionally offer them lower-paid trips.
Vulnerable Immigrant Workforce: Most gig workers are immigrants (many from authoritarian countries) who fear legal trouble or job loss, making them hesitant to demand their democratic and data rights.

4 Surveillance and Control: Which Actors Gain Power?

📍 Case Study II: Offline Tracking (Aweso Mowlana)

A 5-year Uber driver (4.95 rating) had the Uber app on two phones (Samsung & iPhone). While working in South London with his Samsung, his son accidentally took his iPhone to Uxbridge. Uber’s bot detected two locations, accused him of "account sharing," fired him, and got his license revoked. WIE recovered his Driver Detailed Device Data and found Uber recorded 230 rows of data per minute (GPS, battery, speed, IMEI) from his iPhone—even though driver_online was "FALSE" (Offline) all day!

🤖 Case Study III: "Fraud" as Control (Alexandru)

Bolt invested €150M in anti-fraud AI, Ola uses "Guardian," and Free Now used "Random Forest" fraud scores to allocate work. Uber driver Alexandru received automated termination threats for "fraudulent activity." A 25-minute support call revealed the bot flagged him simply for taking a short detour around roadworks and rejecting low-paid trips! Three months later, Uber admitted it was a "glitch." Platforms also use Behavioural Nudging (Uber Pro points) to force an 88% acceptance rate and late-night weekend shifts.

Which Actors Gain Power in this "Surveillance Arms Race"?
1. Platform Companies: Gain total monopoly control over labor without employee liabilities.
2. Regulators (e.g., TfL): Transport for London blindly trusts Uber’s faulty algorithm reports to revoke licenses and demands monthly firing reports.
3. Police & Security Services: London Met Police sends 2,000+ data requests/year to Uber (vs. 411 across all of Canada). Intelligence Case Study: An innocent driver was suspended for 7 weeks (losing £5,000) after Uber mistakenly gave his name and data to police investigating "drug supply" instead of the actual suspect's name.

5 Transparency and Accountability

GDPR Rights vs. "Black-Box" Algorithms: EU GDPR grants workers three core rights: Article 15 (right to access data and processing logic), Article 20 (right to machine-readable CSV data portability), and Article 22 (protection against purely automated firing without human intervention).
Dark Patterns & DSAR Obstruction: WIE submitted 500+ DSARs across 7 apps and exposed systematic obstruction (only Deliveroo cooperated properly; Bolt claimed sharing route data was "commercially devastating"; Amazon Flex & Just Eat sent messy PDFs; Uber stopped sharing telematics data in 2020):
- Circular Answers (M. Ahmed): Uber deleted his account, then bots repeatedly told him to "log into your deleted account" to request data.
- Hidden Data (Mr. Amini): Uber gave 30 days of empty post-firing data, then hid 32 out of 50 data fields in his geolocation file.
- Resistance (Mr. Majid): After a month of delays, Uber told a London driver to travel to Amsterdam in person to hand-deliver a physical letter!
Fake Human Review (Rubber-stamping) & UK Threats: To bypass Article 22, Uber claims humans review every firing, yet admitted in Aweso Mowlana's case that staff blindly approve ("rubber-stamp") bot decisions. Meanwhile, the UK government proposed weakening Article 22 and adding data request fees.

6 Participation & Inclusion + 7 Technology and Society (SCOT)

Exclusion & Co-management: In the "fuzzy front end" of design, only bosses, developers, and regulators (TfL) are included; drivers are excluded ( ruining lives and causing conspiracy theories among workers). MEP Leïla Chaibi calls for algorithmic "Co-management" by unions alongside WIE's Worker Data Trust.
Technological Determinism vs. Social Constructivism (SCOT - Pinch & Bijker, 1984): Platforms claim automated bots are the only way to manage millions of rides (Determinism). However, this report proves Social Constructivism ($Technology \leftrightarrow Society$). Under Interpretive Flexibility, different Relevant Social Groups view the exact same algorithm differently:
• Platform Owners: A cost-saving, law-evading business "Solution."
• Regulators (TfL) & Police: A crime-fighting surveillance "Tool."
• Gig Workers: A racist, livelihood-destroying "Problem."
How Society Shapes Technology Back (Legal Victories): Just as women's movements reshaped bicycle design in our lecture, workers and WIE are forcing platforms to change their algorithms through law: Italy fined Deliveroo/Glovo; Spain passed Europe's first workplace AI transparency law; and in March 2021, the Amsterdam District Court delivered historic rulings in Uber I, Uber II, and Ola—ruling Ola's automated pay cuts illegal under GDPR Article 22, forcing Ola to reveal its "Guardian" and fraud score logic, and reinstating 6 robo-fired Uber drivers with compensation.

8 What Can We Learn? (Key Takeaways for Conclusion)

1. Data Rights = Labor Rights

Workers controlled by data cannot defend their labor rights without data transparency. As MEP Kim van Sparrentak said, "Knowledge is power"—we need 21st-century laws to end "Computer says no."

2. Strict Enforcement Needed

Winning "Limb (b) worker" status in UK courts failed to stop robo firing. As MP Andy McDonald noted, we must end bogus self-employment and grant full employment rights from day one.

3. Collective Data Trusts

Individual court fights are too slow and costly for poor drivers. Pooling data via Worker Data Trusts (WIE) is the most effective weapon for collective bargaining.

⏱️ 12-Minute Presentation Slide Plan (4 Minutes Per Member)

Speaker & Time	Slides	Core Topics & Case Studies Covered
Member 1 (First 4 Minutes) Questions 1 & 2	Slide 1: Title Slide 2: Context Slide 3: Themes	• Introduction to Case Study 1.6 ("Managed by Bots" by WIE). • UK Gig Economy (4.4M workers), Misclassification 2.0, and London firing stats (10,169 fired / +123%). • Course Themes: Datafication of Work, Algorithmic Management & 3 Data Types (Input, Observed, Inferred).
Member 2 (Middle 4 Minutes) Questions 3 & 4	Slide 4: Bias Slide 5: Tracking Slide 6: Power	• Technology is Political & Racial Bias: Pa Edrissa Manjang’s Facial Recognition Case (94% BAME drivers). • Ubiquitous Surveillance: Aweso Mowlana (230 rows/min offline tracking) & Alexandru (Anti-fraud bot glitch). • Who Gains Power: Platforms, Regulators (TfL), and Police Intelligence Sharing (7-week wrongful suspension).
Member 3 (Final 4 Minutes) Questions 5, 6, 7 & 8	Slide 7: GDPR Slide 8: SCOT Slide 9: Lessons	• Transparency Crisis: GDPR (Articles 15, 20, 22), Dark Patterns & 3 DSAR Cases (Ahmed, Amini, Majid). • Inclusion (Co-management), SCOT Framework (Interpretive Flexibility) & Amsterdam Court Rulings (Uber/Ola). • Conclusion: Inseparability of Data & Labor Rights, Full Employment Enforcement, and Worker Data Trusts.

"Managed by Bots": Robo Firing in the Gig Economy