Key Takeaways: Chapter 28 — Algorithmic Management

Core Arguments

1. Algorithmic management is the computational culmination of scientific management's century-old project. Taylor's goal was the elimination of worker discretion through systematic measurement and direction. Algorithmic management achieves this at a scale, speed, and granularity Taylor could not have imagined — but the structural relationship is identical. The algorithm is Taylor's foreman, running continuously at millisecond resolution, without fatigue, judgment, or relationship.

2. The key distinction from monitoring is control: algorithmic management doesn't just measure behavior, it directs it. Performance monitoring creates data for human review. Algorithmic management uses data to issue real-time directives, generate automated discipline, and make termination recommendations — without continuous human decision-making. Jordan cannot ask the algorithm why they received a specific task assignment because the algorithm doesn't take questions.

3. Amazon's fulfillment center is the world's largest deployment of algorithmic management, and its consequences are well-documented. The rate metric, automated warnings, TOT tracking, and documented "fired by algorithm" cases at Amazon provide the most detailed empirical picture of what large-scale algorithmic management looks like in practice. The bathroom break controversy crystallizes the system's inability to distinguish constrained from chosen behavior.

4. Gig economy platforms apply algorithmic management without employment protections. Uber, Lyft, DoorDash, and similar platforms exercise comprehensive algorithmic control over workers — task allocation, performance monitoring, behavioral nudging, deactivation — while denying employment status. The misclassification allows platforms to have the control of employment without the legal obligations of employment.

5. The black box manager problem creates a structural accountability gap. When consequential decisions are made by opaque algorithms, there is no human who made the decision, no supervisor who can explain it, no HR representative who reviewed the individual case. Workers who want to contest algorithmic decisions have nowhere to take their grievance. The distributed responsibility problem means no one is accountable because the algorithm processed it.

6. Customer rating systems transform customers into unwitting distributed surveillance agents. The five-star rating, experienced by customers as a simple feedback mechanism, is experienced by gig workers as a consequential performance evaluation that can trigger deactivation. It is affected by factors beyond workers' control and cannot be contested by affected workers.

7. Algorithmic management systems encode management values in their design — they are not neutral. The performance score formula, the idle time threshold, the task assignment logic — these encode choices about what to optimize for, whose interests to center, and what counts as performance. The Python simulation in this chapter makes these embedded values explicit in a way that real systems deliberately obscure.

8. Emotional labor monitoring through sentiment analysis extends algorithmic measurement into the affective interior of work. Sentiment analysis of customer calls measures not just what workers say but the emotional register of how they say it — extending the algorithmic gaze into domains of experience that workers have historically retained some autonomy over, even in intensively monitored environments.

9. Worker organizing against algorithms requires new strategies adapted to algorithmic conditions. The ALU at JFK8 and the RWDSU at Bessemer both demonstrate that traditional organizing approaches require adaptation for algorithmic management contexts: encrypted communications for organizing discussions, demands for algorithmic due process (not just wages), and legal strategies targeting algorithmic unfair labor practices.

10. International legal frameworks are developing faster than U.S. frameworks. The EU's Platform Work Directive, AI Act, and GDPR Article 22 create a substantially more protective framework for workers subject to algorithmic management than anything in U.S. federal law. The gap between U.S. and EU worker protections in this area is significant and growing.

Essential Vocabulary

The Five "Cans" of Algorithmic Due Process

Workers subject to algorithmic management should ask: 1. Can I see the performance data being used to evaluate me? 2. Can I understand the algorithm's methodology — what it measures, how it weights factors, what thresholds trigger consequences? 3. Can I contest specific automated decisions with a human reviewer? 4. Can I know what the algorithm does and does not account for (disability, facility conditions, etc.)? 5. Can I organize without the monitoring infrastructure being used against my protected activity?

In most current algorithmic management systems, the answer to all five is no or severely limited. These are the demands that labor organizing in the algorithmic era is beginning to articulate.

Connections to Recurring Themes

Visibility asymmetry: The algorithm knows everything about Jordan; Jordan knows almost nothing about the algorithm. This is visibility asymmetry at its extreme — the asymmetry is not just between supervisor and worker but between an omniscient system and a worker who cannot even see the system's interface.

Consent as fiction: Jordan agreed to work in a facility with a warehouse management system. They did not agree to specific thresholds, assignment logic, or automated discipline procedures — details that were neither disclosed nor consented to in any meaningful sense.

Normalization of monitoring: Algorithmic management is presented as efficient, objective, and neutral — making it difficult to perceive as surveillance. "The system assigned that" is already a normalized response to what was, structurally, a consequential management decision.

Structural vs. individual explanations: The bathroom break controversy is the paradigm case: the individual worker's choice to urinate in a bottle is a rational response to a structural constraint created by algorithmic design. The structural analysis locates the problem in the design; the individual analysis blames the worker.

Historical continuity: Algorithmic management is Taylor plus computation. The ideology (eliminate worker discretion through systematic measurement and direction), the asymmetry (management's science vs. workers' bodies), and the consequences (higher rates, more discipline, higher turnover) are continuous with the scientific management era.

Looking Ahead

Chapter 29 turns to the hiring process — examining how the same algorithmic logic that manages existing workers is applied before anyone is hired. HR analytics, AI-powered video interviews, resume screening algorithms, and predictive hiring tools extend the algorithmic gaze to the moment before the employment relationship begins — and Jordan's internship application provides the most direct connection to their own life circumstances.