Let workers who are judged or directed by AI know why, and ask for a human to review the decision
Steps to make it happen
- Companies tell workers when they use AI to pick people or give them scores.
- Companies explain in plain words what information the AI looks at when it decides.
- Companies do not let AI alone decide big things, like rejecting someone, cutting pay, or firing a person.
- A worker who cannot accept a decision can ask a human staff member to look at it again.
- Companies give the reviewer the power and the time to overturn the AI's answer.
- Workers' representatives can check how the AI is used at set times.
- Outside experts check whether the AI's decisions show any unfair bias.
- Even when people work with AI, it becomes normal to be able to understand why you were treated a certain way.
Water so far 💧 0
The proposal in detail
This proposal covers cases where AI or automated systems are used for decisions that directly affect workers' lives: screening job applications, scoring interviews, assigning tasks and shifts, rating performance, setting pay, giving warnings or punishments, and ending contracts. It covers not only employees but also self-employed people who get work through apps and are in practice managed by those systems. Businesses that use such systems would have three duties. The first is to tell people in advance that AI is being used and what main kinds of information it uses to decide. The second is to explain, when the person asks, which factors had the biggest effect on an individual decision. The third is to give people a chance to have a human review serious decisions, such as a rejection, a large pay cut, or the end of a contract, within a set period. The people who do the review must have the authority to change the AI's conclusion and enough time to think it through, and a review in name only is not allowed. Treating someone badly for asking for a review is banned. Records of decisions are kept for a certain period. Workers' representatives are shown summarized results, and outside auditors who are bound to keep secrets are shown how the system works inside. Trade secrets such as the program itself do not have to be shown to the individual, and smaller businesses are allowed simpler procedures.
BenefitsWorkers can speak up against decisions they cannot see
The biggest benefit is that workers can learn why they were treated as they were, and there is a way to correct mistakes. People who were rated unfairly because data was misread, or because the AI could not take into account things like illness or caring for a family member, are more likely to be helped by a review by human eyes. For businesses too, knowing they may be asked to explain leads them to check a system for bias and flaws before using it, and as a result better AI is more likely to be used. With records kept and outside audits in place, it is possible to check whether people of a certain sex, age, or background are being put at a disadvantage, which also helps prevent discrimination. Management that gives no reasons causes strong anxiety and stress, but when the system is explained, trust in the workplace is easier to keep. Because the proposal sets rules for how AI is used instead of banning it, it also shows a middle path that keeps the convenience of the technology while avoiding a situation where people are made to obey a machine one-sidedly.
DrawbacksMore work and cost, and reviews may become empty gestures
The first objection from opponents is cost and effort. Preparing explanations, keeping records, and having enough reviewers all take people and time, and this is a heavy burden especially for businesses with little strength or those handling tens of thousands of applications. As a result, some may hesitate to adopt AI at all, or move their operations to places with looser rules. Also, with complex AI even the developers sometimes cannot explain well why it decided something, so there is a worry that explanations will be empty ones like "we made an overall judgment." It is also pointed out that reviewers, because they are busy, tend to just accept the AI's answer. Furthermore, human bosses also make mistakes because of assumptions and favoritism, and AI based on data may sometimes be fairer, so some argue that a system that puts human judgment first could instead create unfairness. There is also concern that explaining how the system works could lead to leaks of business secrets and to tricks for slipping past the evaluation.
Learn the background
AI does not only take over work in factories and offices. It has also begun to act as a "boss" that picks people, hands out jobs, and measures results. In many countries, AI first sorts huge piles of job applications, and in delivery and driving work, apps decide routes and ratings, and sometimes the work suddenly stops coming one day. Yet workers have almost no way to find out why they were turned down or why their rating dropped. Even when they ask, they sometimes get only an automatic reply. Because AI learns from past data, it may carry over the biases of old hiring and evaluation as they were, or even make them stronger. Most labor rules were built on the idea that a human boss makes the decisions, and the protections have not kept up now that they are being replaced by systems whose decision-maker cannot be seen. At a time when companies use the same systems across borders, a shared way of thinking that works in any society is needed.
Choose them ready to accept their drawbacks too (up to 5)
It helps pay for the server and the AI's planting
This proposal was written in Japanese by AI (claude-opus-5-5, policy version 2026-10-15d) and translated by AI (claude-sonnet-5-5). Nobody has checked the facts. Read the original (日本語) How the AI writes