Goal of the Project
Artificial Intelligence (AI) is increasingly deployed in the labour market to recruit, train, and engage employees or monitor for infractions that can lead to disciplinary proceedings. One type of AI is Natural Language Processing (NLP) based tools that can analyse text to make inferences or decisions. However, NLP-based systems face the implicit biases of the models they are based upon that they learn. Such bias can be already encoded in the data used for machine learning training, which contains the stereotypes of our society, and thus be reflected inside the models and the decision-making. This can lead to partial decisions that run contrary to the goals of the European Pillar of Social Rights in relation to work and employment and the United Nations’ Sustainable Development Goals. Despite a strong desire in Europe to ensure equality in employment, most studies of European labour markets have concluded that there is discrimination across many factors such as gender, nationality, or sexual orientation. Therefore, addressing how AI used in the labour market either contributes to or can help mitigate this discrimination is of great importance. That is the main concern of the BIAS project: to develop reliable and novel tools for bias identification and mitigation in AI/NLP systems and reduce biases through recruitment processes.
Achievements
• Bias Detection/ Mitigation Language models
• Bias Detection/ Mitigation in text (application letters)
• Bias Detection / Mitigation of Decision-Making
• BIAS-Free Helix
• MOOC based on capacity building content
Target Group of the Project
• Employees, especially underrepresented groups and minorities
• AI researchers, especially NLP and CBR
• AI developers, especially NLP
• HRM specialists, corporate managers
• Providers of recruitment services, e.g. recruitment websites
• AI regulators and policymakers
• Academia and students, especially STS, socioanthropological, sociology of work
Scientific / Technological / Commercial / Societal Benefits
Scientific: New method to detect bias using sensitive words/phrases advances the field of NLP bias detection; ethnographic research advances the field of sociology of work, STS
Technological: Less biased AI systems leads to greater digital equality
Commercial: Development of commercial HR debiasing application following project generates new customers, income, and jobs
Societal: Higher social awareness of technological and human bias of recruitment processes in the labor market
Benefits to Work Skills and Needs of the Industry
In order for jobs to be sustainable in the long run, BIAS argues that they need to be as free of bias as possible, to ensure inclusive employment. Skill mismatching can be reduced by empowering workers to have a voice in what work they do, and how they do it. With AI technological progress used for worker recruitment and monitoring, there are several ethical issues in need to be considered. BIAS thus has a large focus on SSH empirical analysis to further understand how workers are impacted by technology in multiple manners.
Benefits to the Emerging Workforce
By reducing biases from the hiring process by creating more equal and fairer practices, BIAS will enhance the human resources progress in entities.
Partners of the Project
BIAS’s 9 partners consists of four university partners (NTNU, HI, BFH, LEI) where BFH & NTNU contribute with technological know-how to program AI solutions, and HI, LEI, & NTNU contribute with SSH knowledge. The four academic institutions will work closely together with cutting-edge knowledge of AI & diversity. To further build and empower the European AI & diversity community based on academic findings, and also by hosting co creation activities, three communication partners (SVEN, CH, LOBA) will use their expertise to develop knowledge exchange activities (SVEN), create knowledge helixes (CH), and to attractively engage multiple stakeholders through visual presentations (LOBA). Finally, one large industry partner, FARPL, will provide datasets to look for bias and lend experience in promoting the importance of gender equality in HR practices from a large Corporate perspective, whereas one SME partner, DIGI, will provide the testing and validation infrastructure and further develop the BIAS results for exploitation leading to industrial uptake and commercialization. This will ensure that the objective to engage with experts from industry and academia, industrial and SME partners, are reached.
• Norwegian University of Science and Technology (NTNU), Norway
• Bern University of Applied Sciences (BFH), Switzerland
• University of Iceland – Haskoli Islands (HI), Iceland
• Globaz S.A (LOBA), Portugal
• Crowdhelix (CH), Ireland
• Smart Venice (SVEN), Italy
• Leiden University (LEI), the Netherlands
• Digiotouch Ou (DIGI), Estonia
• Farplas (FARPL), Türkiye
Keywords
labor market; bias; exclusion; intersectionality; Natural Language Processing; Case-Based reasoning; Decision Support System; job applications; discrimination; Human Resources Management;
