2026/ Case file P03 / Draft
> CASE FILE 2026 ............ OPEN
> SUBJECT ........ ONE DRIVER, BEIJING
> SOURCE ......... ONE INTERVIEW, ~10 MIN
> STATUS ......... DRAFT / PENDING REVIEW

2026

The algorithm is watching you

That line is a provocation, borrowed from the world of Orwell’s Nineteen Eighty-Four — not a finding. What follows is built from one recorded conversation with a ride-hailing driver in Beijing, the coding of that conversation, and outside research. Every piece is labelled for what it is.

Q · Who is doing the watching?

Not this page. It sets no cookies, loads nothing from anyone else, and keeps no record of you. The eye follows your pointer and forgets it instantly. The only thing under observation here is the evidence.

Begin the file

Chapter 01 · Interview testimony

A day begins
at 07:00

Before the system speaks, a person does.

On 25 October 2026, in a ride-hailing car in Beijing — the navigation voice running in the background — a student researcher talked with the driver about his work.

P03 · recording metadata

On when his day starts

“Around seven in the morning.”

Translated from Mandarin E01 · S2 · T002.01 · 00:00:13

Ten seconds later — on how long he works

“About eleven hours.”

Translated from Mandarin E02 · S2 · T004 · 00:00:23

Approximate, and his own. It says nothing about how long any other driver works.

That is all this archive knows about the shape of his day: a start, and a rough length.

Everything in between is unrecorded. No trip log, no earnings record, no GPS trace — and we will not invent one.

One more fixed point exists, and it belongs to us, not him: the interview itself was recorded from about 17:51 to 18:02.

P03 · metadata · GMT+08

On what is hardest about the work

“Not making money is the hardest part.”

Translated from Mandarin E03 · code IV01 · locator pending

A statement of experience, not earnings data. No income figures appear anywhere in this pamphlet.

Chapter 02 · Simulation · Pending editorial review

An order
arrives

Here the pamphlet stops quoting and starts imitating.

The phone is an invented, generic interface — not modelled on any real app, carrying no real data. No route, fare, rating or platform name appears, because none is sourced. It exists only to put a few seconds of work in your hands.

Simulation

  1. Go online to begin. Without JavaScript, read on below.

The driver’s word for how orders reach him

“Compulsory assignment.”

Translated from Mandarin E06 · code IV06 · locator pending

And on the route

“Navigation isn’t up to the driver.”

Translated from Mandarin · his words repeat and hesitate; the hesitation is kept in the archive E05 · code IV05 · locator pending

These describe how the work feels to him. They are not a specification of any dispatch system. Whether, when and at what cost an order can be declined is something one interview cannot establish — which is exactly what the buttons on the phone make look simple.

You are offline

New request

Pickup
not simulated
Drop-off
not simulated
Fare
no data

Route set

Follow the line.

End of simulation

Nothing you did here was recorded.

SIMULATED INTERFACE · NOT A REAL PLATFORM

Chapter 03 · Testimony, interpretation, literature, absence

Your score /
who decides?

A dystopian story would show you a dashboard here. We tried to build one.

Commission rate

—

No record

Account score

—

No record

Penalties

—

No record

No platform records available

The numbers spin because a dashboard expects numbers. They land on nothing because this archive has none.

On payment

“They take a cut on every ride.”

Translated from Mandarin E04 · code IV04 · locator pending

Every ticket below loses a piece. How big a piece, we cannot say: no audited commission rate exists in the evidence, so each cut is drawn with a question mark instead of a size.

Read the testimony four ways

A · What was said

Both translated from Mandarin. Neither comes with a number.

Self-report

B · How it was coded

In the revised mixed coding these phrases were tagged as In Vivo codes IV04 and IV09 — labels that stay close to the speaker’s own words. A code is an act of reading: it marks a passage as meaningful; it does not verify it. Method

The coded document holds 91 unique codes. That number describes our analysis, not how common any experience is among drivers. P03

Interpretation

C · What research says

Scholars use algorithmic management for systems in which software allocates tasks, monitors work, ranks workers and applies incentives or discipline; a 2026 systematic review organises the field along those lines. Frontiers 2026

A UN DESA policy brief uses similar vocabulary for platform work worldwide, and a 2026 study looks at how ride-hailing workers respond to algorithmic management. UN DESA 2026 Shen et al. 2026

The literature gives us words for what he describes. It does not prove his platform works the way any study says.

External

D · What we don’t know

  • What share of each fare is deducted, or how it is calculated.
  • Which actions lead to points or fines, and whether the rules are published.
  • Whether he has been penalised, or is describing a risk he feels.
  • The exact transcript turns for E03–E07.
  • Whether he consents to these excerpts being published.
Unknown

Chapter 04 · From the class manuscript · Open question

Not every
driver agrees

It would be easy to end here, with one voice confirming the poster. The research behind this pamphlet doesn’t allow it.

The team’s draft manuscript reports that the drivers it discussed did not describe platform work in one way. Turn the cards. P01 · qualitative, not quantified

Figures are abstract patterns, not portraits. No other driver quotations exist in this archive.

Why might the same work feel different?

The manuscript suggests three factors. A hypothesis, not a result.

  • Prior employment
  • Income needs
  • Expectations

Request refused. Three positions are not one finding — and there are no percentages here until comparable, consented interviews exist.

So the question is not “how bad is it?” but for whom, compared with what — and who gets to say?

Chapter 05 · Theory, not finding

Who serves
whom?

Frame one

A centaur: a person uses a machine to extend what they can do. The person directs; the tool assists.

Frame two

Cory Doctorow uses reverse centaur for workplaces in which a human is fitted around a machine. The system directs; the person supplies. Doctorow 2025

Frame three

A different idea — Clark and Chalmers’ extended mind — argues that thinking can reach into the tools a person relies on. The class manuscript draws on it. Clark & Chalmers 1998

When is a navigation app part of how a driver thinks — and when is it something that thinks for him?

Set against “compulsory assignment” and “navigation isn’t up to the driver,” the frames pose a question, not an answer.

In a given hour of a given day, which picture is closer? Only the person driving can say — and he might say different things on different days.

These frames are not the same, and neither is evidence.

Chapter 06 · Archive and accountability

Open
the files

The poster at the top borrowed the authority of an institution. This chapter hands it back: here is where every claim came from, and where the evidence runs out.

  1. E01Self-report

    “Around seven in the morning.”

    Locator
    P03 · S2 · T002.01 · 00:00:13
    Gloss
    Starts taking orders around 7 a.m.
    Caution
    No independent time log.
  2. E02Self-report

    “About eleven hours.”

    Locator
    P03 · S2 · T004 · 00:00:23
    Gloss
    An approximate length of his working day.
    Caution
    Approximate; not a claim about other drivers.
  3. E03Locator pending

    “Not making money is the hardest part.”

    Locator
    P03 · code IV01 · turn to be confirmed
    Caution
    Subjective experience, not earnings data.
  4. E04Locator pending

    “They take a cut on every ride.”

    Locator
    P03 · code IV04 · turn to be confirmed
    Caution
    Perceived commission; no audited rate.
  5. E05Locator pending

    “Navigation isn’t up to the driver.”

    Locator
    P03 · code IV05 · turn to be confirmed
    Caution
    Hesitation in the original is preserved in the archive; does not prove full automation.
  6. E06Locator pending

    “Compulsory assignment.”

    Locator
    P03 · code IV06 · turn to be confirmed
    Caution
    The driver’s label, not a product specification.
  7. E07Locator pending

    “Deduct points and fine you.”

    Locator
    P03 · code IV09 · turn to be confirmed
    Caution
    A felt risk; no confirmed penalty rule.

All quotations are working English translations. The Mandarin originals are kept in the project’s research archive, not on this page.

Sources held by the project
P01
Collaborative class manuscript, The research of internet-drivers, Platform Labor, and Algorithmic Management. Working research; not independent proof of platform mechanisms.
P02
Line-by-line In Vivo coding of the interview, with English labels translated from the Mandarin source.
P03
Revised mixed line-by-line coding with transcript appendix. Recorded 25 October 2026, approx. 17:51–18:02 (GMT+08); 110 timestamped turns; 140 meaning units; 91 unique codes; 133 code applications. These counts describe the analysis, not a sample.
P04
Earlier coding draft, superseded by P03. Not used.

None of these files, and no audio, is published here.

Method · four ways of coding
In Vivo
Keeps the participant’s own expression. The Mandarin original is the evidence; English is a translation.
Etic
A researcher’s theoretical category. Never a quotation.
Process
Labels what the speaker describes doing or undergoing.
Versus
Marks contrasts visible in the conversation, with context kept.

No code frequency in this project should be read as how common an experience is among drivers.

External references
  1. Doctorow, C. (2025). Reverse centaurs are the answer to the AI paradox. Original. Interpretive lens only.
  2. Clark, A. & Chalmers, D. (1998). The Extended Mind. Analysis 58(1). doi:10.1093/analys/58.1.7
  3. Frontiers in Sociology (2026). Systematic review of algorithmic management. Article. A taxonomy, not proof about any one platform.
  4. Shen et al. (2026). Algorithm aversion among ride-hailing workers. Article.
  5. UN DESA (2026). Policy Brief No. 194: Decent work in the platform economy. Brief. Global policy vocabulary.
  6. Rest of World (2022). Millions of workers are managed by algorithms. Feature. Other workers, other places.
  7. Human Rights Watch (2025). The Gig Trap. Report. United States only.
  8. Orwell, G. (1949). Nineteen Eighty-Four. Referenced as atmosphere; not quoted.
Limits and open questions
  • One fully coded interview: a single recording of about ten minutes. It represents one person on one day.
  • Statements about commissions, penalties, navigation and dispatch describe experience; they do not verify any platform’s rules.
  • E03–E07 must be matched to exact transcript turns before publication.
  • Consent to publish excerpts and translations is not yet confirmed.
  • No platform is named, by default.
  • Research from other countries is context, not evidence about Beijing.
This page and you

No cookies, analytics, fingerprinting, third-party fonts or scripts. The eye in the opening reads your pointer position to draw itself and stores nothing. The simulation keeps no record of your choices. The slogan is fiction; the absence of tracking is not.

The file stays open.