With great power comes great responsibility.
HRTECH has a curious relationship with law and regulations. On the one hand, it owes its genesis to payroll. Vendors have built sophisticated methods to gather and code these legislative requirements. They work closely with governments to understand what laws are coming, and how to code them into the software. The vendors understand the differences between laws in different countries, and either specialise, or build products that aid compliance in multiple jurisdictions. Often payroll vendors have expertise in pay related law that rivals that of top lawyers and the regulators themselves. Lawmakers and vendors generally collaborate well.
At the same time, HRTECH seeks to help employers optimize processes such as recruitment, performance appraisal, internal mobility, and succession planning. The industry makes all sorts of promises about the power of AI in this regard. However, many of these products are built with scant regard to elements of national labour, data protection and anti-discrimination laws. Software built in the US is often simply sold in Europe and elsewhere, with not a moment’s thought of how the application interacts with the laws elsewhere.
For GDPR, for instance, vendors tend to rely on consent, which is usually pretty meaningless in the employment relationship context. Freely given isn’t freely given when the alternative is unemployment. Informed isn’t informed if the processing is a mystery.
A recent academic paper by Lilian Edwards, Lina Dencik and Javier Sánchez-Monedero examines the issues of how US-made algorithmic hiring systems may cater to US anti-discrimination law but fail to meet EU standards.
Our new paper " (@LinaDencik , @lilianedwards and Javier Sánchez-Monedero) is up! We argue that US-made algorithmic hiring systems may cater to US anti-discrimination law but fail to meet EU standards esp re DP. This time w link 🙂 https://t.co/Tvg5Gghdzn— Lilian Edwards (@lilianedwards) October 2, 2019
Herewith the punchline.
Connectedly , the transfer of AHSs (AHS is the term the authors use for automated hiring systems) developed within a US socio- legal context to a UK (and arguably EU) context introduces a number of fundamental legal problems of fit, not just with regards to discrimination and equality law, but as we have argued, perhaps more significantly in relation to DP law. GDPR transparency rights in arts 15 and 22 may provide avenues to overturn aspects of the candidate-employer information asymmetry and might even outright prohibit the use of AHSs for wholly automated decision- making in hiring. Yet these rights may be ignored or ill- implemented in systems not built within the EU.
There a myriad reasons for this curious state of affairs, where for some laws HRTECH provides and enables compliance, and for others ignores or perhaps actively circumvents. I’ve spent time exploring this in my own academic endeavours, and I should really turn the 400 pages of my diss into more bite size pieces.
For this state of affairs to change, the behaviour of the software buyer must change. If you start asking vendors about how their products comply with GDPR, and don’t just take the consent tick box on face value, that would give them pause. For instance, ask probing questions about how product design looked at the the UK Equality Act. Ask them to explain how their lack of accessibility features squares up to the European Web Accessibility Directive and the European Accessibility Act. If you ask better questions, you will get better answers, and better products.
As we apply AI to HR, we are impacting the very fundamentals of the right to work, of equality and fairness. Think back to that time you were looking for a job, and think for a moment how you would like software to treat you in that process. I suspect you would want that software to have been designed to take into account the laws of the land in which you live, and treat you fairly and with dignity. Ask your vendors how they do that.
To the software builders, at the risk of repeating myself, I point you to the words of one of the pioneers of AI, Joseph Weizenbaum.
The computer programmer is a creator of universes for which he alone is the lawgiver. No playwright, no stage director, no emperor, however powerful, has ever exercised such absolute authority to arrange a stage or field of battle and to command such unswervingly dutiful actors or troops.
On a related point, I’ll be picking up the topic of accessibility in the recruitment process in my talk at Unleash Paris in a couple of weeks time. You can get a discount here. I’m speaking on day 2 at 15:00 in Room W04.
I'm a venture capitalist at Acadian Ventures, investing in the future of work.