Ups and Downs of Human Resources Management Algorithms
Introduction to three use cases of HRM algorithms in public and private sectors
Written by Aurore Guglielmi, Ethel Tan, Yuxi Wang and Ricardo Zapata
The digital transformation of certain sectors have been more rapid than others. Human resources management as a business function concerning the allocation of talent and labour touches on all sectors and has received significant attention of late.
According to IBM’s Global C-suite Study in 2018, 65% of CEOs surveyed across 112 countries expect skills management to have a significant impact on their businesses and require the creation of new strategies. The awareness is present. However, only 28% expect real action to be taken by their companies. This gap between expectations and reaction can be attributed to a host of factors — ranging from the changes and size of needs, the level and adoption of technology, the agility and inertia of actors, and the various legal, sociopolitical and economic consequences.
This work studies three cases of deploying artificial intelligence (AI) and algorithms in human resources management across the public and private sectors: Buona Scuola in Italy, Amazon in the US, and Precire in Germany. We will understand the forces pushing for and pushing against the incorporation of new technologies in human resources management.
To put this into context, we need to keep in mind that the stage and speed of adoption is still nascent but fast. IBM’s Global C-suite Study in 2018 found that most companies are struggling to build analytics capabilities in human resources. 41% of CEOs are not prepared to use data analytic tools and only 4% of CEOs are prepared to a large extent. Furthermore, a survey by LinkedIn in 2018 saw that less than one-quarter companies in North America have already adopted analytics inhuman resources. However, the need is strong, due to (1) competition for a limited talent pool; (2) complexities in human resources management (e.g. diversity, geolocation, hiring, benchmarking, branding); (3) quick development of human resources management innovation and technology. The nascent stage but fast speed of adoption creates a tension. There is considerable uncertainty and controversy involved — which will be illustrated in the case studies.
Nonetheless, a quick analysis of opportunities and challenges shows that there is still great potential to be tapped on to improve the processes and outcomes of human resources management. The capability of AI to conduct high-speed computation, stable and predictive analytics, and accuracy with quality data, is able to manage the heavy paperwork, repetitive administrative work, inherent psychology and bias involved in human resources management. We can end up with reduced decision-making bias, reduced administrative time, have access to better insights, improved strategic hiring and training and a personalised employee experience. However, many issues have yet to be properly addressed at a technology-, corporate policy-, and public policy-level — such as privacy and security, fairness, small data sets, integration and maintenance of the technology, accountability, liability, resistance to adoption. By looking at the following three case studies, we aim to address the challenges where applicable and identify best practices.
CASE I: BUONA SCUOLA | Italy
“The algorithm that decided the destiny of many families”
The Buona Scuola case deals with the poor implementation of simple allocation algorithm in the midst of a hectic political context. As part of a teachers’ mobility program in 2016, an algorithm was supposed to assign vacant positions across the country to existent and new teachers. After its implementation, thousands of teachers found to be wrongly placed. News coverage began to unveil the personal drama of separated families and the missing explanations on the side of the Ministry of Education. The case was taken to courts, teachers were reassigned to new posts, and just until early-2019 Italy’s State Council ruled creating valuable jurisprudence in favour of higher algorithmic accountability. This case is relevant as it brings transparent and explainable automated public decisions into the centre of debates around digital human rights.
Check here the complete story.
CASE II: AMAZON | United States
Discriminatory AI hiring system
Amazon’s experimental hiring tool used artificial intelligence to give job candidates scores ranging from one to five stars — much like shoppers rate products on Amazon.” Amazon’s team has been building machine learning program to review job applications since 2014. However, the system did not rate candidates for software developer jobs and other technical positions in a gender-neutral way. The data being fed into the system come from successful applicants to Amazon over a 10-year period. Unfortunately, the data set is biased against female applicants because there is an overwhelming majority of male dominance across the tech industry especially for technical jobs.
CASE III: PRECIRE | Germany
Precire is a new speech analysis software particularly dedicated to Human Resources. It helps recruiters assess stress levels and psychological stability of candidates. The software analyses your voice digitally, looking at your voice variation, vocabulary and syntax. It will therefore be able to determine your personality, your strengths and weakness. Based on its finding, it will provide recruiters with diagrams and charts summarizing those traits. The form is therefore more important than the content. In the case of Precire, the software runs without any human intervention. It is supposed to be able to determine ‘artificial emotional intelligence’ more accurately than any human. For instance, use of words such as partly or slightly will make the candidate stand out as insecure.
LESSONS LEARNED & CONCLUSION
AI in human resources management can have a strong potential to raise employee engagement and productivity. It is essential to combine (1) people, (2) process and (3) technology to deliver transformational value at an optimised cost. Employees are most affected so it is crucial to focus on their needs and outcomes.
This post was written for the “Data & Algorithms for Public Policy” class instructed by Simon Chignard, Timothée GIDOIN and Jean-marie JOHN MATHEWS at the Sciences Po Paris School of Public Affairs as part of the “Digital, New Technology & Public Policy” master policy stream courses.
