abusesaffiliationarrow-downarrow-leftarrow-rightarrow-upattack-typeburgerchevron-downchevron-leftchevron-rightchevron-upClock iconclosedeletedevelopment-povertydiscriminationdollardownloademailenvironmentexternal-linkfacebookfiltergenderglobegroupshealthC4067174-3DD9-4B9E-AD64-284FDAAE6338@1xinformation-outlineinformationinstagraminvestment-trade-globalisationissueslabourlanguagesShapeCombined Shapeline, chart, up, arrow, graphLinkedInlocationmap-pinminusnewsorganisationotheroverviewpluspreviewArtboard 185profilerefreshIconnewssearchsecurityPathStock downStock steadyStock uptagticktooltiptwitteruniversalityweb

Cette page n’est pas disponible en Français et est affichée en English

Article

30 Déc 2021

Auteur:
Bengo4.com

Japan: First labour dispute involving IBM over AI-driven wage assessments continues for more than 18 months after IBM allegedly refuses to disclose information requested by union

"「AIによる賃金提案」に上司は支配されるのか 新たなIBM労使紛争で見えたこと" 30 December 2021

[Japanese-to-English translation: Business & Human Rights Resource Centre]

Japan's first labour-management dispute over AI-driven wage assessments and personnel evaluations is pending before the Tokyo Metropolitan Labour Relations Commission. In April 2020, the IBM Japan branch of the JMITU, a labour union of IBM Japan employees, filed a petition for relief with the Commission, claiming that the company's refusal to engage in collective bargaining in a good faith over AI-driven personnel assessments and wage reviews constituted an unjustified refusal to bargain and intervention in control.

The Commission is still investigating the matter after more than 18 months since the filing of the complaint. Masafumi Hozumi, a lawyer representing the union, pointed out that, "IBM needs to provide explanations in a good faith when introducing AI-driven wage assessments if the union demanded explanations of the AI's learning data and output. Refusing to do so and not coming to the negotiating table is a violation of the right to organise and bargain collectively recognised in Article 28 of the Constitution and the Trade Union Law.” IBM, on the other hand, says the AI-driven wage assessment is "just a tool to support managers’ decisions."

[...]

[...]According to documents provided by IBM, IBM Compensation Advisor with Watson is "an AI-driven system that provides information for salary adjustments" and considers "40 different types of data" in order to make salary increase recommendations.

Examples include "the amount of skill in the market", "the expertise of the main job", "the need for skills at IBM" and "past salary increases".

These will be assessed for each of the four factors of "skills", "competitiveness of base salary", and "performance and career potential" before a salary proposal is made.

[...]

Hozumi says "we are not denying the use of AI in personnel evaluation." However, he points out that there are four potential disadvantages for workers that are caused specifically by AI.

  1. Invasion of privacy
    Employees' attributes and daily behaviour may be monitored and, in some cases, information that should not be collected may be gathered. There are concerns that the collection and use of information other than personal performance and job performance may infringe on workers' privacy. IBM Japan has not disclosed in collective bargaining what information it allows Watson to take into account.
  2. Issues of fairness and discrimination
    There is a concern that those who speak and act in an affirmative manner towards those who are in a dominant position in the company will be highly valued, while those who do the opposite will be devalued. Incidentally, in the case of IBM Japan, union members are a minority, representing only a few percent of the total workforce. In the past, they have fought against the company to withdraw wage cuts due to power harassment. While justice and ethics may change over time, AI does not have justice or ethics and cannot recognise and correct discrimination on its own. Discrimination against those who belong to the minority side can be magnified and reproduced.
  3. Black boxing
    AI cannot determine what is right and cannot explain the process that led to the decision. Employees who receive a low evaluation cannot even improve themselves without knowing why they were given a low evaluation. Opportunities for workers to develop can be lost.
  4. Automation bias (the tendency to overestimate computerised, automated decisions)
    An IBM union member was once told by a supervisor that "Watson wanted to give you a raise, so we gave you one". A sales presentation on Watson used by the company also includes statements to the effect that managers are subject to Watson's decisions. IBM Japan positions Watson as a tool to "support" HR evaluation, but the automation bias is likely to make it impossible for managers to defy the AI.

[...]

The point of this complaint is about how workers' wages are determined. Article 28 of the Constitution of Japan and the Trade Union Act recognise three rights for workers: the right to organise, the right to collective bargaining and the right to collective action.

Hozumi said: "If the company is going to introduce AI, we should discuss in collective bargaining what the AI will do, what algorithms it will use, what data it will take into account, what it will output, how the managers will use it, and so on, and if there are any improvements that need to be made, we should make them, and as far as possible reach an agreement. That's how it should be done. IBM Japan's refusal to even come to the bargaining table is a violation of the Constitution and the Trade Union Act."

[...]

The timetable for the future is that the Commission is expected to examine witnesses in the spring of 2022.