Platforms and Gig Work
平台与零工
This paper asks what a person who takes jobs through a platform is treated as by the company. In 2023 the Amsterdam Court of Appeal heard the case of four ride-hailing drivers against Uber: their accounts had been deactivated for "suspected fraud," and for three of them Uber could not show that there had been actual human intervention before the deactivation. In the app such a person is a user; in the contract he is often written as a partner; in fact his work is dispatched by the platform, his price set by the platform, his ratings kept by the platform, and his account deactivated by the platform. This paper does not argue for a half-way legal status. It says one thing: he actually stands in several relations, and the requirements of each of those relations all apply in his case. Identification takes two steps: first, whether he is in the interior, by Institution Theory's two branches, of which the first is enough; then, whether the platform actually organizes and controls what he does, by what is actual, not by the name in the contract. From this: whoever takes an employer's control owes an employer's share of protection. Dispatching, pricing, rating and deactivation are decisions that fall at positions: the reasons can be given, they can be questioned, and questioning has actual efficacy. Pricing by the job does not in itself cross the line; it is cost pressed into the Base Layer that crosses it; for the cost of waiting online, the normative layer first asks who bears it, and measurement is left to the rule layer. For jobs and time slots he has not yet taken on and that do not fall within established work duties, his refusal or choosing otherwise does not bring on adverse arrangements to the point that he cannot in fact refuse; what the platform calls an arrangement does not replace the identification of work duties; and whether the track record can be taken along is one of the places asked about first. The paper lays down three propositions (PT1 to PT3), written out at the ends of the sections and listed together at the end.
Writing Declaration: This paper was co-drafted with Claude (Anthropic). All intellectual decisions, framework design, and final editorial judgments were made by the author.
Introduction
On April 4, 2023, the Amsterdam Court of Appeal gave judgment in the case of four ride-hailing drivers against Uber (ECLI:NL:GHAMS:2023:793). Three of the four lived in the United Kingdom and one in Portugal; all of their Uber accounts had been deactivated on the ground of "suspected fraud."
Uber told the court that the decisions to deactivate accounts were taken by employees of its risk team, at least two of them, and unanimously. For three of the drivers the court found that Uber had not shown actual human intervention before the deactivation. Each of several employees had left a note, but the notes "tonen geenszins aan dat die handeling veel meer is geweest dan een louter symbolische handeling" (in no way show that that act was much more than a merely symbolic act; my translation). The court also wrote: "Het is moeilijk voor te stellen hoe dat, vanuit Krakau, kan worden gedaan, zonder dat de chauffeurs daarbij worden gehoord." (It is difficult to imagine how that can be done, from Krakow, without the drivers being heard; my translation.) The remaining driver had had an interview with Uber before the deactivation, a fact that was not disputed; on that basis the court held that actual human intervention had been sufficiently shown.
Only this one practice is taken here, together with the court's finding on it. Later in this paper two further findings about Uber are used, each stated with its year and the practice concerned: what is judged is the practice, and no company is characterized as a whole.
From this the question: what are these four people treated as by the platform? In the app they are registered users. In the contract they are often written as partners, as independent contractors. In fact their work is dispatched by the platform, their price set by the platform, their ratings kept by the platform, and their accounts deactivated by the platform.
There is a saying: such a user is really half an employee. This paper does not argue for a half-way legal status, nor does it decide his legal standing for any jurisdiction. What it says is this: he actually stands in several relations, and the requirements that arise from each of those actual relations all apply in his case; the contract writes only one of them, and the others do not lapse on that account. Nor does this mean that registering for an app stacks on one person all the rights of a consumer and of an employee under some jurisdiction. It is about where the requirements come from: from the actual relation, not from the name.
§1 and §2 do the identification: whether he is in the interior, and whether the platform actually organizes and controls what he does. §3 takes up algorithmic management; §4, cost; §5, leaving and choosing; §6, watching and asking; §7, direction.
The reading follows the Prequel: the construct takes the ought, the position takes accountability, the person takes his own law.
§1 Is He in the Interior?
What a company owes to whom begins with who is in the company's interior. Institution Theory has laid down: "The interior is drawn by entanglement: those in the entanglement the construct handles, together with those its acts reach who have no way to avoid it by exiting, make up the interior; reach of influence is not reach of authority, and whether the construct is organizing their lives is judged by the actual relation." (Institution Theory Z4)
The two branches stand side by side; belonging to either one puts a person in the interior. Institution Theory Paper 1 says it plainly: "People in the first part are not necessarily unable to leave"; "Only the two together make the interior; neither part is the whole."
A person who takes jobs on a platform falls in the first branch. Dispatching falls on him, prices fall on him, ratings fall on him, and deactivation falls on him too: he is directly within the entanglement the platform handles. To recognize him as in the interior, the first branch is enough; there is no need to prove further that he cannot leave. Someone who takes a job now and then and has a full-time job elsewhere, and someone who lives by it, are both in the interior.
For those who already fall in the first branch, the difficulty of leaving does not decide whether they are in the interior; it decides only the degree of the relation. Paper 3, §1 wrote: "The harder it is to leave, the more explanation and routes of questioning the company owes him"; "the ease of leaving does not cancel his standing to have the line guarded and to question." The original speaks of users, and leaving there means switching to another provider; Paper 1 has a sentence of the same form for investors (GS4), and Paper 4 for partners (GS10). Here it is extended to the person who does work on a platform, for whom leaving means leaving a relation he relies on to work and to live (extension; own undertaking). Applied here, it is a baseline with increments above it: the reasons can be given, they can be questioned, and the channel of questioning is effective, the same for everyone; the harder it is to leave, the more explanation, routes of questioning and checks the platform owes him above the baseline.
The difficulty of leaving counts only the part that can be examined: whether ratings and job records can be taken along; whether taking jobs on another platform at the same time is penalized; how much has been invested in a vehicle and equipment; what share of the local market the platform holds. Market share is one clue in identifying the difficulty of leaving and says nothing on its own; it must also be examined whether he can move to work elsewhere, what switching costs, and how income and track record are affected. The harder it actually is to get around, the closer the platform is to what Political Economy calls a chokepoint, "a position others cannot get around on the way to sustaining their lives or unfolding their own paths" (Political Economy Paper 2, E7). Whether in his heart he can bear to leave is his own affair, and is not counted in the degree.
§2 Does the Platform Actually Organize and Control What He Does?
What the first step gives. He is in the interior, so Institution Theory's five oughts fall on the platform, each on its own conditions. The five are: "not crossing the line, guarding the line, providing conditions for cultivation, letting cultivation between persons happen, and remaining open to questioning" (Institution Theory Paper 0). The most direct for him are not crossing the line (Institution Theory Z9) and remaining open to questioning (Institution Theory Z34).
Institution Theory also wrote a sentence that speaks of exactly his kind of person: "For internal subjects in the entanglement who hold no registered position at the construct, the ought does not take hold through their positions; it takes hold through the construct's constraints on itself and through arrangements, with a source, at other positions within the construct; the five oughts apply each on its own conditions and are not reduced for want of a position." (Institution Theory Z4) The contract does not give him an employee's position; the five oughts are not reduced all the same.
What the first step still lacks. These are addressed to the construct, and do not yet directly give the three propositions of Paper 2 (GS5 to GS7). Those three stand on the relation of working together: the company buys what he does, and he works within work the company organizes. "A company buys what is done, not people" is a constraint on the company once this relation holds; it does not on its own carry the identification of the relation: the company of Paper 4 also buys what partners do, and a company that buys results from an independent supplier is likewise buying "what is done."
So the second step asks not only whether money was paid and what result was bought, but how the platform actually organizes his work: who sets the price, who sets the terms, who decides dispatch and cancellation, who keeps the ratings and what they are used for. The latter half of that sentence of Institution Theory has already given the reading: "whether the construct is organizing their lives is judged by the actual relation." (Institution Theory Z4) Once this step holds, the three propositions of Paper 2 apply.
The name in the contract does not decide. Under Gate One (E21), Political Economy Paper 4 quotes a sentence from Rights Theory Paper 5: "The criterion lies in structure, not in professed voluntariness." There it speaks of the consent signed on taking a job; here the subject term is changed to the name the contract gives the relation: whether it is called partner, independent contractor or user, the name does not decide the reading on its own (own undertaking). Paper 3 has a sentence of the same form, "A model is judged by its execution, not by its name" (GS9), which can be read alongside: GS9 gives the reading that a name does not judge in place of the facts; Z4 gives the ground of the interior. That one and the same person is not missed in any of his several relations is this paper's extension, its own undertaking.
Under any of the ways of writing it, his protections are no fewer. Papers 2, 3 and 4 each laid down protections for one relation: employees (GS5 to GS7), users (GS8, GS9), partners (GS10). He actually stands in the first two relations, yet the contract often writes him into the third. Even taking the contract at its word and treating him as a partner, that proposition of Paper 4 falls on him: "A person has no equivalent, all the way down the chain" (GS10). The person on the partner's side is himself.
A precedent. On February 19, 2021, the UK Supreme Court gave judgment in Uber v Aslam ([2021] UKSC 5): the drivers are workers (a category in English law between employees and independent contractors). The court said that taking the written agreement as the starting point is "wrong in principle." It listed five points at which Uber controls the drivers: the fare is set by Uber; the contractual terms are set by Uber; accepting and cancelling trips is constrained by Uber; ratings are kept by Uber, which can terminate on that basis; and communication between driver and passenger is restricted by Uber.
These five points help identification, but they are not a checklist, and the finding is not complete as soon as one of them appears. The judgment weighed the relations as a whole. At paragraphs 103 and 104 it compared accommodation booking platforms specifically: there too are standard terms, commission and ratings, yet the accommodation is not a standardized product defined by the platform, "it is, crucially, the supplier and not the platform which sets the price," and the ratings do not serve as "a system of internal performance measurement and control." With the same contracts, charges and ratings, the actual relation can differ.
This precedent lies at the rule layer. How legal status is recognized, each jurisdiction has its own route: Article 5 of the EU Platform Work Directive (October 23, 2024; Member States must transpose it by December 2, 2026) presumes an employment relationship on finding "facts indicating direction and control," with the burden of rebuttal on the platform; Spain's Law 12/2021 presumes an employment relationship between delivery platforms that organize, direct and control through algorithmic management and their riders; the United Kingdom has the intermediate category of worker; California's Proposition 22 of 2020 creates a separate set of benefits without recognizing employment, and on July 25, 2024 the California Supreme Court upheld it. A presumption of employment (the EU, Spain), an intermediate category (the UK), separate benefits (California): the three routes land in different places. The normative layer does not choose among them; it asks only three things: whether his protections in each of his relations are all there; whether algorithmic management can be questioned; whether the track record can be taken along.
After identification: whoever takes an employer's control owes an employer's share of protection. A platform cannot hold an employer's control with one hand and a contractor's immunity with the other.
If the platform declines the corresponding protections of working together on the ground that he is free to take jobs and free to go online, then his refusal of jobs and time slots he has not yet taken on and that do not fall within established work duties cannot bring on adverse arrangements to the point that he cannot in fact refuse. What already falls within shared work duties with an identifiable source is examined separately under §5.
If the platform uses downgrading, account suspension or dispatch arrangements to compel him to keep taking jobs, the three propositions of Paper 2 must be examined by this actual relation of organizing and controlling work.
Dispatch arrangements made because of his current unavailability, an unsuitable distance, or work he has already undertaken do not by themselves warrant that inference; they are examined separately under §5.
The point in Aslam about "accepting and cancelling trips" falls exactly here.
PT1 Status is judged by what is actual, not by its name. One person is at once a user of the platform and someone who does work on it, and is sometimes written by the contract as a partner; the requirements that arise from each of these actual relations all apply in his case, and the name in the contract reduces none of them; the five oughts are not reduced for his want of a position. If the platform declines the corresponding protections on the ground that he is free to take jobs and to go online, then his refusal of jobs and time slots he has not yet taken on and that do not fall within established work duties cannot bring on adverse arrangements to the point that he cannot in fact refuse; what already falls within shared work duties with an identifiable source is examined separately by authority, performance, the protection of others, and questionability (per Institution Theory Z4; extended from Political Economy E21; the same form as GS9; own undertaking).
§3 Algorithmic Management Is Management
Dispatching, pricing, rating and deactivation are all decisions. That a decision is output by a system does not change that it is a decision: it falls at a position; someone set up the system, someone set its parameters, someone decided to use it. A decision does not lose its place of explanation and of being questioned because a system output it.
The reasons can be given, they can be questioned, and questioning has actual efficacy. The first step already gives everyone in the interior a channel of questioning: the construct's ought to remain open to questioning is "not reduced for want of a position" under Z4 (Institution Theory Z34). Where the second step holds, it is examined further by what Paper 2 wrote of employees' voice: "On decisions that affect employees, employees can ask through a channel independent of management, and asking has actual efficacy" (GS7). Paper 6 wrote of judging hard times: "The reasons can be given and can be questioned" (GS13). Here the latter sentence is extended to everyday decisions such as dispatching and deactivation (extension; own undertaking). Applied to algorithmic management, the channel of questioning is examined by Political Economy's three questions: "who can activate it, what actual efficacy it has, and how the party questioned cannot cancel it on its own" (Political Economy Paper 3, E16).
Deactivation weighs most. Deactivating a person's account cuts off his entrance to work and income on this platform; when he lives by it, the effect is especially heavy. An emergency temporary suspension and a final deactivation must be kept apart: where a passenger is endangered, the platform may suspend first; explanation, the chance to answer, and substantive review must still be secured in good time, and are not waived because the suspension came first.
Attribution falls on the construct. Where a platform deploys an algorithm, the address of attribution is the construct that deploys it; the platform cannot present the algorithm as another address. Political Economy wrote: "The construct is the address of attribution; the construct's account, the position's accountability and individual responsibility with independent grounds are examined separately: the construct's name does not shield the last, and the construct's failure is not automatically transferred onto any one person" (Political Economy Paper 1, E2). How this lands once execution is handed to AI is handed to this series' paper The Company in the Age of AI.
The rules a platform makes. If a platform actually becomes an entrance to work and income that many people find hard to get around, the ways it sets dispatching, pricing, rating and appeals fall on many people's way to sustaining their lives. Of the rules made by a position others cannot get around, the Specialist Paper wrote: "The scope of these rules is drawn by the institution, not fixed and changed by the rule-making party on its own"; "The closer to a chokepoint, the stronger this requirement" (A Way Out, a Way Forward, E41a). Whether a platform is close to a chokepoint depends on whether those who work there actually have another way (Political Economy Paper 2, E7); the determination is handed to the theory of the market. How the scope is drawn, and by which level of institution, lies at the rule layer.
Examples at the rule layer. Explaining first, the chance to answer, review by someone with authority to change the decision: their order and form lie at the rule layer (GS7: "its form is left to the rule layer"). Examples that can be given:
Article 10(5) of the Platform Work Directive: a decision to restrict, suspend or terminate the contractual relationship or the account "shall be taken by a human being."
Article 11 of the same Directive: platform workers have the right to an explanation of automated decisions; the platform must designate a contact person with the necessary competence, training and authority to discuss and clarify the facts and reasons behind a decision.
Article 22 of the EU General Data Protection Regulation: a person has the right not to be subject to a decision based "solely on automated processing" that produces legal effects or similarly significant effects; it also lists exceptions, and lays down safeguards according to the kind of exception: for the two kinds of contractual necessity and explicit consent, at least the right to obtain human intervention, to express one's point of view and to contest the decision; for the kind authorized by law, suitable safeguards laid down by the law relied on.
Two findings. In the case of the Introduction, when the court examined the deactivation decisions, it cited three items from the guidelines of the European Data Protection Board: oversight must be meaningful and not a merely symbolic act; it must be carried out by someone with the authority and the competence to change the decision; and all relevant data must be taken into the analysis. For three of the drivers, Uber's employee notes did not suffice to show these three; for the other, the fact that there had been an interview before the deactivation was not disputed, and on that basis the court held that actual human intervention had been sufficiently shown. The side this paper takes from it is this: deactivation is a management decision affecting the entrance to work; someone must be able to give the reasons and be questioned; whether there was actual intervention must be examined in the process of that particular handling, whether all relevant data were looked at, whether he was heard, whether the one taking this step had the authority and the competence to change it; it cannot be settled by counting how many people appeared in the process.
The other is a regulatory finding. According to the European Data Protection Board's announcement of October 8, 2026, the Dutch Data Protection Authority fined Uber 824,990,000 euros for a practice of 2018 to 2022: where fraud was suspected or passenger ratings were low, drivers' accounts were automatically deactivated, some temporarily and, for persistently low ratings, permanently; these were fully automated decisions, and the information given about profiling was also insufficient. The announcement records that "Uber has now stopped the violations"; Uber has appealed the fine, and as of the announcement there was no final judicial decision. Only the practice in this one finding is taken here; the outcome of the appeal will govern later.
PT2 Algorithmic management is management. Dispatching, pricing, rating and deactivation are decisions that fall at positions: the reasons can be given, they can be questioned, and questioning has actual efficacy; a decision does not lose its place of explanation and of being questioned because a system output it. The address of attribution is the construct (per Institution Theory Z34, by way of Z4, Political Economy E2, E16; per GS7 where the second step holds; extended from GS13; own undertaking).
§4 The Cost Falls on What Is Done
The main line of Paper 2 is: a company buys what is done, not people; the price falls on what is done, not on the person. Wages are the price of what people do; "a company may adjust what people do, but does not trade off the person himself" (GS5). Applied to pricing by the job, several cases must be kept apart.
Pricing by the job does not in itself cross the line. One price per job is one way of pricing what is done. More jobs or fewer, higher prices or lower, can be negotiated, agreed, and adjusted to circumstances.
What is pressed into the Base Layer crosses the line. Waiting for jobs, driving empty, platform price changes, the slow season: if these costs are pressed down layer by layer to the layer by which he sustains his life, they are first examined by this series: the cost falls on what is done, and the person's Base Layer is not used to offset it (GS5); Paper 4, on partners, puts it more directly: "it does not press costs, layer by layer, into the Base Layer of the people on the partners' side" (GS10). If the platform goes further and uses his dependence for his livelihood to make refusal or exit fail in fact, it runs into Political Economy's Gate One: what people need to sustain their lives is not converted into control that compresses their life-substrate (Political Economy Paper 4, E21).
Clawing back money already earned. If the platform afterwards deducts money he has already earned, and the deduction is tied to his leaving, it falls in the case Political Economy Paper 2 describes: what is tied to "not going" is not an additional payment but "wages already earned, deferred pay, or money paid out and then clawed back" (Political Economy Paper 2, E12a). If it is tied to refusing jobs, it is read by the latter half of Gate One: "nor may money, housing or the manner of paying wages be used to place people's ability to leave, refuse or revoke a standing bond under another party's control" (Political Economy Paper 4, E21). Both press him with what is his own, and the choice is false.
Other clawbacks, corrections and refunds are tendencies at the rule layer. Paper 2 wrote: "Sharing first, progressive sharing, and not taking away what has already been earned are tendencies at the rule layer" (GS5). Refunding a fare paid in error and recovering money obtained through proven fraud each have their own grounds and are handled at the rule layer.
Waiting online: the norm asks first, the rules measure. He has the app open, at the designated place, waiting for a job. Does this time count as what he does?
The normative layer does not decide in advance for every logged-in state, nor does it push the whole question onto the rule layer. It first asks three things: whether the platform obtains from his being online what it wants, namely capacity on call; whether the platform makes his being online a condition of dispatch; and who bears the cost of this availability. Where a platform actually requires or controls a person's keeping himself available for it, the cost of standing by does not automatically fall to zero because there is no order yet. Which periods count, how they are calculated and how they are shared are handed to the rule layer.
At the rule layer there are two examples, pointing in different directions. In Aslam, the court upheld the employment tribunal's finding that drivers logged into the app within the authorized territory and able and willing to accept trips were working (paragraph 123); being logged into other platforms at the same time is judged by fact and degree, and in that case there was no evidence that drivers could in practice be at the disposal of other operators at the same time (paragraphs 134 to 137). California's Proposition 22 builds its earnings floor for each earnings period from 120 percent of the applicable minimum wage for "engaged time" (the time from accepting a request to completing it), plus compensation for vehicle expenses calculated on the miles of engaged trips; where net earnings fall below that total, the platform makes up the difference. The time spent waiting online is not counted in its time component. The same Proposition's occupational accident insurance, by contrast, covers the time he is online. The normative layer does not directly prescribe a single method of measurement; each method must still be examined for how the cost of the availability the platform requires of him is shared.
§5 A Way Out, a Way Forward
The Political Economy Specialist Paper A Way Out, a Way Forward puts these two things first at the normative layer: "In reading an arrangement, the normative layer first asks two things: whether there is a way out, and whether there is a way forward." (A Way Out, a Way Forward, E40) Applied to platforms, two kinds of affairs must first be kept apart.
The entrances he decides himself. For jobs and time slots he has not yet taken on and that do not fall within established work duties, whether to take them and whether to enter is his to decide. Whether to use another platform, whether to seek another livelihood, is likewise his to choose; how these connect with shared arrangements already formed is examined below. The Specialist Paper lists leaving, refusing and where to go as a person's own affairs (A Way Out, a Way Forward, E41).
Shared arrangements already formed. How a job already accepted is completed, how a time slot already committed to is coordinated, whether taking jobs elsewhere leaves him unable to complete what he has already undertaken, the limited conditions needed to protect passengers and to complete particular tasks: these are shared affairs. The Specialist Paper wrote: "In the affairs he shares with others, such as the arrangement of work, the scope of authority is drawn by the institution and can be questioned, and whether the line is crossed is examined by exercise" (A Way Out, a Way Forward, E41). Shared affairs too do not pass automatically on "already agreed" alone; their authority, scope, actual effect and questionability are still examined.
The consequences of refusal are examined by the arrangement of work that actually holds.
For jobs and time slots he has not yet taken on and that do not fall within established work duties, his refusal or choosing otherwise must be practically possible: the platform cannot use downgrading, account suspension or dispatch arrangements to make this choice fail, to the point that he cannot in fact refuse. Adverse arrangements brought on by refusal are at the same time evidence of the control described in §2. Arrangements made because he is unavailable at the moment, too far away, or must first complete what he has already undertaken are not of this kind, and are examined separately by actual effect.
If the platform uses downgrading, account suspension or dispatch arrangements to compel him to keep taking jobs, the three propositions of Paper 2 are examined, as in §2, by the actual organizing and controlling of what he does.
When a task already falls within shared work duties with an identifiable source, the refusal itself is still his; its consequences are examined by authority, performance, the protection of others and questionability, the three propositions of Paper 2 apply, and the restriction reaches only that one matter. Shifts, on-call periods or post duties already established can also include tasks not yet claimed one by one; what must be examined is the source and scope of those duties. What the platform calls the arrangement does not replace this identification: a platform's acknowledging that it organized the work can be evidence for identifying the relation, but cannot by that create its authority over any particular task.
"Retaliation" is not the word here. Institution Theory Paper 2 wrote: "Crossing the line does not require that anyone intend it" (Institution Theory Z10). To count as a limit, a constraint must meet the conditions set out in Z9, including not sealing off the channels of exit, refusal and revocation; "one that crosses any of these is a violation" (Institution Theory Z9). So there is no need first to prove that anyone meant to punish; one looks only at what has structurally fallen on him after the refusal. The relation between the adverse arrangement and the refusal must still be examined; that it came after the refusal does not by itself complete this identification.
Taking jobs on several platforms. Political Economy Paper 2 judged non-compete clauses: after a person leaves, they write a part of his space of directions as not to be chosen, which is violation (Political Economy Paper 2, E12a). Working for another at the same time during the term of an agreement, and where one goes after leaving, are two different matters. Here it is examined along two routes.
Whether it obstructs shared work he has already undertaken is examined by the authority, scope, performance and questionability of shared affairs (A Way Out, a Way Forward, E41).
Whether a payment is used to tie up his choice is examined by the three criteria of conditional payment in Political Economy Paper 2, taken together: "is the backing prohibition and compulsory enforcement, or only no longer paying; is what is tied to 'not going' an additional payment, or wages already earned, deferred pay, or money paid out and then clawed back; taken together with other constraints, or under deep dependence, does refusing the payment amount to being unable to live; the three are used together, and that the backing is only no longer paying is not yet enough to recognize that the choice is actually there" (Political Economy Paper 2, E12a).
Read in this way: if what happens is only that an additional bonus not yet earned is no longer paid, and, taking the three together, a real choice remains after refusing, the normative layer does not judge it a strict violation, and the rule layer may permit it or not (Political Economy Paper 2, E12a). Cutting into the dispatch he lives by, or combining with "ratings cannot be taken along" to seal off his leaving, crosses the line. Bonuses, deductions and the allocation of dispatch opportunities are all examined by actual effect: a bonus too can lock a person in; fewer dispatches can also come from distance and availability, and need not be a penalty (extended from Political Economy E12a; own undertaking).
The track record. If the job records and ratings a person has built up on one platform cannot be carried over when he moves to another, he must start from scratch. Paper 3 wrote: "Whether data can be taken along, and whether unsubscribing is no harder than subscribing, are the two places the normative layer asks about first" (GS8). The original speaks of users' data; extended to the person who does the work, whether the track record can be taken along is also one of the places asked about first. If it cannot be taken along, leaving is hard; the harder it is to leave, the more explanation, routes of questioning and checks the platform owes him (extended from GS8; own undertaking). "Can be taken along" itself is taken as a direction, and as a tendency at the rule layer.
The rule layer already has provisions. Article 9(6) of the Platform Work Directive: platform workers have a right to the portability of personal data generated through their platform work, "including ratings and reviews," without prejudice to the rights of recipients of the service under the General Data Protection Regulation; the platform must provide tools free of charge and, at his request, transmit the data directly to a third party.
The judgment layer must keep three things apart: whether his own work records can be obtained and exported; whether exporting them would also disclose private data belonging to third parties such as passengers; and how a new platform interprets and gives weight to these records. Portability does not require another platform to accept the original scores or weights. Ideally, exported records include the explanations needed to understand them, such as the categories of tasks, the basis of the ratings and the status of dispute handling, without including third parties' private data.
PT3 Refusing new jobs and choosing otherwise must be effective; the track record is asked about first. For jobs and time slots he has not yet taken on and that do not fall within established work duties, his refusal or choosing otherwise does not bring on adverse arrangements to the point that he cannot in fact refuse; when a task already falls within shared work duties with an identifiable source, the consequences of refusal are examined by authority, performance, the protection of others and questionability; what the platform calls the arrangement does not replace this identification. Whether the track record can be taken along is one of the places asked about first; if it cannot, the platform owes him more explanation, routes of questioning and checks (per A Way Out, a Way Forward, E40, E41, Institution Theory Z9, Z10, Political Economy E12a; extended from GS8; own undertaking).
§6 Watching and Asking
Watching: monitoring reaches only the work, not the person. Not crossing the line is given already by the first step (Institution Theory Z9); where the second step holds, it is examined further by that sentence of Paper 2 (GS6). While he is online and taking jobs, his driving route and delivery times are work; where he went after going offline, whom he talked to and what he said, how he felt, are not.
Article 7(1) of the Platform Work Directive writes several of these as prohibitions: platforms may not use automated systems to process data on the emotional or psychological state of platform workers, may not process private conversations, may not collect personal data while he is not offering or performing platform work, and may not use data to predict whether he will exercise the right of association or take part in collective bargaining. This is one example at the rule layer; the normative line follows Institution Theory Z9 and GS6, and does not stretch or shrink with the scope of any one law.
Asking: the channel must be real. On the rules and decisions that affect him, he can ask through a channel independent of the platform's operating side, and asking has actual efficacy: what the first step gives is Institution Theory Z34; where the second step holds, it is examined further by GS7; and "those who ask are not penalized for asking" (GS12). If the algorithm changes in a way that affects his income, he can ask. It is examined by the three questions: who can activate it, what actual efficacy it has, and whether the party questioned can cancel the channel on its own (Political Economy Paper 3, E16).
Examples at the rule layer: Spain's Law 12/2021 added an item to Article 64.4 of the Workers' Statute giving workers' representatives the right to be informed of "los parámetros, reglas e instrucciones en los que se basan los algoritmos" (the parameters, rules and instructions on which the algorithms are based; my translation); Article 10(1) of the Platform Work Directive requires platforms, with the involvement of workers' representatives, to evaluate regularly the impact of automated decisions on people.
The user's side. He is at the same time a user of the app. No locking in, no manipulation, no deception follow Paper 3 (GS8, GS9), and are not rewritten in this paper.
§7 Direction: Giving Him Conditions for His Work
The reasons for dispatching and pricing can be given, the decisions that affect him can be questioned, and questioning has actual efficacy; jobs and time slots he has not yet taken on and that do not fall within established work duties can be refused; whether the track record can be taken along is asked about first. These are requirements already laid down above, not direction.
Above these lies direction. Contribution is a direction, not a duty; outcomes are not taken as rules. A platform oriented toward the Kingdom of Ends can give those who work on it more: letting people know earlier before the rules change, fuller information for anticipating income, ways of working that are easier for him to arrange himself, help in taking his track record to another platform. These give him more of a way forward while he works on the platform, and more room to carry his own path on.
The closer a platform is to a chokepoint, the stronger the requirement that algorithmic management be open to questioning; at the normative layer this has been seen in §3 (A Way Out, a Way Forward, E41a); how it is strengthened accordingly, and how the track record is made easier to take along, lie at the rule layer.
These are directions, not labels. This paper does not characterize any platform; a platform not oriented toward the Kingdom of Ends is bound all the same by the lines of the two gates, a way out, and openness to questioning.
Where This Paper Stops
This paper has laid down three propositions: status is judged by what is actual, not by its name (PT1); algorithmic management is management (PT2); refusing new jobs and choosing otherwise must be effective, and the track record is asked about first (PT3).
What it has not proved: how the legal status of platform workers should be legislated; social insurance contributions and tax; competition between platforms, self-preferencing and mergers; what pricing is reasonable.
What it hands elsewhere: the determination of platform competition and chokepoints, to a later theory of the market; the design of classification and social insurance, which lies at the rule layer, to each institution and each level of government; once execution is handed to AI, what conditions the position of judgment must meet and how attribution lands, to The Company in the Age of AI. That paper and this one are the two ends of one and the same chain of decision: in the case of the Introduction, the deactivated drivers are the ones decided upon, and this paper writes of them; the employees placed in the reviewing position, that paper writes of them.
Remainders left: between someone who takes a job now and then and someone who lives by it, there is no ready-made line; the baseline is the same for everyone, the strength above it varies with the difficulty of leaving, and how it increases is examined item by item (§1). Which periods of waiting online count and how they are counted, the normative layer has given only three questions (§4). Which reductions in dispatch after a refusal were brought on by the refusal and which come from distance and availability depends on actual effect, and there is no general cut; which tasks already fall within work duties with an identifiable source is also examined item by item (§5).
The Three Propositions of This Paper
PT1 Status is judged by what is actual, not by its name. One person is at once a user of the platform and someone who does work on it, and is sometimes written by the contract as a partner; the requirements that arise from each of these actual relations all apply in his case, and the name in the contract reduces none of them; the five oughts are not reduced for his want of a position. If the platform declines the corresponding protections on the ground that he is free to take jobs and to go online, then his refusal of jobs and time slots he has not yet taken on and that do not fall within established work duties cannot bring on adverse arrangements to the point that he cannot in fact refuse; what already falls within shared work duties with an identifiable source is examined separately by authority, performance, the protection of others, and questionability (per Institution Theory Z4; extended from Political Economy E21; the same form as GS9; own undertaking).
PT2 Algorithmic management is management. Dispatching, pricing, rating and deactivation are decisions that fall at positions: the reasons can be given, they can be questioned, and questioning has actual efficacy; a decision does not lose its place of explanation and of being questioned because a system output it. The address of attribution is the construct (per Institution Theory Z34, by way of Z4, Political Economy E2, E16; per GS7 where the second step holds; extended from GS13; own undertaking).
PT3 Refusing new jobs and choosing otherwise must be effective; the track record is asked about first. For jobs and time slots he has not yet taken on and that do not fall within established work duties, his refusal or choosing otherwise does not bring on adverse arrangements to the point that he cannot in fact refuse; when a task already falls within shared work duties with an identifiable source, the consequences of refusal are examined by authority, performance, the protection of others and questionability; what the platform calls the arrangement does not replace this identification. Whether the track record can be taken along is one of the places asked about first; if it cannot, the platform owes him more explanation, routes of questioning and checks (per A Way Out, a Way Forward, E40, E41, Institution Theory Z9, Z10, Political Economy E12a; extended from GS8; own undertaking).
References
All SAE works below are cited by concept DOI, which always resolves to the latest version.
I · This system
Qin, H. (2026). To Whom the Ought Is Addressed (SAE Institution Theory, Paper 1) [Preprint]. Zenodo. https://doi.org/10.5281/zenodo.23269635
Qin, H. (2026). Not Crossing the Line (SAE Institution Theory, Paper 2) [Preprint]. Zenodo. https://doi.org/10.5281/zenodo.23270384
Qin, H. (2026). Open to Questioning (SAE Institution Theory, Paper 6) [Preprint]. Zenodo. https://doi.org/10.5281/zenodo.23271324
Qin, H. (2026). Institution Theory and the Tradition: The Address of the Ought (SAE Institution Theory, Paper 0) [Preprint]. Zenodo. https://doi.org/10.5281/zenodo.23272468
Qin, H. (2026). Accumulation (SAE Political Economy, Paper 1) [Preprint]. Zenodo. https://doi.org/10.5281/zenodo.23874138
Qin, H. (2026). Chokepoints (SAE Political Economy, Paper 2) [Preprint]. Zenodo. https://doi.org/10.5281/zenodo.23874140
Qin, H. (2026). Checks (SAE Political Economy, Paper 3) [Preprint]. Zenodo. https://doi.org/10.5281/zenodo.23874142
Qin, H. (2026). Conversion (SAE Political Economy, Paper 4) [Preprint]. Zenodo. https://doi.org/10.5281/zenodo.23874146
Qin, H. (2026). A Way Out, a Way Forward (SAE Political Economy, Specialist Paper) [Preprint]. Zenodo. https://doi.org/10.5281/zenodo.24021385
Qin, H. (2026). What a Company Is Not (SAE Theory of the Company, Prequel) [Preprint]. Zenodo. https://doi.org/10.5281/zenodo.24045716
Qin, H. (2026). Investment (SAE Theory of the Company, Paper 1) [Preprint]. Zenodo. https://doi.org/10.5281/zenodo.24045721
Qin, H. (2026). Working Together (SAE Theory of the Company, Paper 2) [Preprint]. Zenodo. https://doi.org/10.5281/zenodo.24045723
Qin, H. (2026). Service (SAE Theory of the Company, Paper 3) [Preprint]. Zenodo. https://doi.org/10.5281/zenodo.24045731
Qin, H. (2026). Partners (SAE Theory of the Company, Paper 4) [Preprint]. Zenodo. https://doi.org/10.5281/zenodo.24045733
Qin, H. (2026). Keeping the Direction (SAE Theory of the Company, Paper 6) [Preprint]. Zenodo. https://doi.org/10.5281/zenodo.24045739
Qin, H. (2026). The Company in the Age of AI (SAE Theory of the Company, Specialist Paper) [Preprint]. Zenodo. https://doi.org/10.5281/zenodo.24117816
II · Judgments and regulatory findings (originals, tier A)
Uber BV and others v Aslam and others [2021] UKSC 5 (19 February 2021), §§76, 83, 103–104, 123, 134–137. https://supremecourt.uk/uploads/uksc_2019_0029_judgment_19c9de2253.pdf
Gerechtshof Amsterdam, 4 april 2023, ECLI:NL:GHAMS:2023:793 (zaaknummer 200.295.742/01), §§3.14, 3.23–3.25. https://uitspraken.rechtspraak.nl/details?id=ECLI:NL:GHAMS:2023:793
European Data Protection Board. (2026, October 8). Dutch DPA fines Uber EUR 824 990 000 for unlawful automated decision-making and insufficient information. https://www.edpb.europa.eu/news/dutch-dpa-fines-uber-eur-824-990-000-for-unlawful-automated-decision-making-and-insufficient_en (appeal pending)
Castellanos v. State of California, S279622 (Cal. July 25, 2024). https://law.justia.com/cases/california/supreme-court/2024/s279622.html
III · Legal Provisions (originals, tier A)
Directive (EU) 2024/2831 of 23 October 2024 on improving working conditions in platform work, Arts. 5(1), 7(1), 9(6), 10(1), 10(5), 11, 29(1). https://eur-lex.europa.eu/legal-content/EN/TXT/HTML/?uri=OJ:L_202402831
Regulation (EU) 2016/679 (General Data Protection Regulation), Art. 22. https://eur-lex.europa.eu/legal-content/EN/TXT/HTML/?uri=CELEX:32016R0679
Ley 12/2021, de 28 de septiembre (BOE-A-2021-15767). https://www.boe.es/buscar/act.php?id=BOE-A-2021-15767
California Secretary of State. (2020). Proposition 22: Text of proposed laws, Bus. & Prof. Code §§ 7453(d), 7455, 7463(j) (as proposed). https://vig.cdn.sos.ca.gov/2020/general/pdf/topl-prop22.pdf (the 2020 text of the proposition; the current code text not checked)
Interfaces
Main upstream: Institution Theory Paper 1, Z4, Paper 2, Z9, Z10, Paper 6, Z34, Paper 0; Political Economy Paper 1, E2, Paper 2, E7, E12a, Paper 3, E16, Paper 4, E21; A Way Out, a Way Forward, E40, E41, E41a; GS5 to GS10, GS12, GS13.
Also used: Paper 3, §1 (the difficulty of leaving and the baseline; the same form as GS4, GS10).
Examples: Uber v Aslam (2021, A); the Amsterdam Court of Appeal judgment ECLI:NL:GHAMS:2023:793 (2023, A); the Dutch Data Protection Authority's fine on Uber (2026, A, appeal pending); Articles 5, 7, 9(6), 10, 11 and 29(1) of the Platform Work Directive, Article 22 of the GDPR, Spain's Law 12/2021 (all A); California's Proposition 22 (the 2020 text of the proposition, A) and Castellanos (2024, A).
Handed downstream: the determination of platform competition and chokepoints, to the theory of the market; the design of classification and social insurance, to the rule layer; the position of judgment and attribution, to The Company in the Age of AI.