Dr Kiki Leutner in conversation with Susan Scott Parker of Disability Ethical? AI
‘Disability in business’ advocates are sounding the alarm on tech enabled recruitment processes. Technology and in particular pre-hire assessments often present a discriminatory hurdle for disabled applicants and mean they miss out on opportunities. To recruiters who typically do not understand disability discrimination the problem is usually much less visible. But they are equally missing out on talent. Companies may offer accessibility features like accessible web sites, but this is not sufficient. Disability equality requires both removing barriers for groups with similar access needs and making reasonable adjustments for individuals to achieve non discriminatory treatment of individuals.
Here, we look at some of the disability equality problems with recruitment technology present today.
1. Disability fair treatment and inclusion is misinterpreted as just being barrier free access for groups.
For example, websites have features that make them easy to read for vision impaired applicants. But individuals have specific needs that might result from a mix of disabilities or personal requirements. These cannot be fully captured by group features. Group based accessibility solutions only ensure access for certain groups and therefore do not meet the criteria for disability equality that must take into account individual needs. Disability inclusion is only possible when the application process does not discriminate against individuals.
2. Assessment processes do not typically declare enough information to applicants
Typically there is no information provided to let applicants a) evaluate whether they are able to meaningfully represent their abilities or personalities during the assessment, b) are able to access the assessment. All assessments should include an accessibility statement to detail the type of assessment and who it might work or not work for. Should applicants decide the assessment would not represent their abilities well, it must be easy for candidates to request adjustments to those assessments in a non-stigmatising process.
3. Offering only one route in, as is typical in recruitment processes today, presents unnecessary barriers.
Employers should avoid the assumption that there is ‘one set of stairs’ applicants must climb. The focus must be on talent rather than random abilities linked to the access requirements of a chosen application process. Employers must offer different application and assessment routes for all applicants. Applicants must be allowed to choose a route that suits them, and be provided with support on which to choose. These routes must be available for all applicants rather than singling out those with disabilities and creating stigma.
People who know that they will need an adjustment or accommodation from employers should be able to request such adjustments early in the application process, having been told what kinds of assessments lie ahead and find out that the company will give it to them at the point of application. Providing different access routes is a first opportunity for employers to do this.
4. Organisations need to develop systems to track and monitor disability and accessibility bias to avoid losing talent.
When using technology for recruitment, disability bias risks being systematised. It is also implemented at a large scale, with many employers using similar systems. Here are some metrics organisations and HR Technology providers may track to quantify disability equality and access:
a. Were all candidates offered the opportunity to ask for an accommodation?
b. How many candidates were offered different forms of assessment? How many different forms of assessment are available and were used?
c. What is the user experience of applicants who get rejected? Are there accessibility or representation concerns?
5. The algorithms used for pre-hire assessments have hidden disability bias, hidden because unlike other biases it is not being tracked with standard fairness measures.
Disabled people are at least twice as likely to be unemployed as non-disabled people. Which means HR algorithms trained with data from people in employment will always show a disability bias. Disabled applicants are different from the normal population, and pre-hire assessments are based on a comparison of applicants against a population norm. As disabled applicants are not sufficiently represented in these populations, algorithms won’t accurately assess them. Fairness is typically evaluated based on group performance. Given that disability cannot be reduced to group factors, this presents a problem for standard adverse impact or fairness evaluations of selection algorithms. Standards and guidelines need to be developed to ensure technology does not result in disability discrimination.
Disability equality in HR Technology (which makes inclusion possible) has a long way to go. In Scott-Parker’s words: “You can’t be included if you can’t get in”. Disability equality lags behind other forms of discrimination like age, gender and ethnicity discrimination, for which practice guidelines are available and fairness testing is implemented as a standard.
New AI Ethics regulation typically extends or supports group based fairness criteria that are insufficient for disability inclusion. As a result, this lack of HR expertise combined with a lack of understanding of disability equality on the part of AI developers threatens the life chances of hundreds of millions world wide. Susan Scott Parker recommends that urgent damage control is needed, and that at the very least AI tools should enable job seekers with tips to navigate AI powered HR platforms – or at least document the experience in ways that make the experience of disabled applicants visible.
What can you do to improve disability equality in hiring?
Helpful Groups and Resources:
https://disabilityethicalai.org
https://www.peatworks.org/employer-topics/artificial-intelligence-ai/
