Improving the Candidate Experience Could Require Lifting the Curtain on AI-Driven Decisions

You may have seen one of the many articles that equate dating with job hunting. Let’s assume that parallel holds true; new research examining the impact of AI algorithms on decision making in the realm of politics and dating align with how employers promote, and candidates evaluate, job opportunities. 

The experiment presented candidates (political and dating) and boosted some over others by suggesting higher compatibility or increasing the number of times certain candidates were presented over others in the set. Results show that participants were influenced differently depending on the type of decision. 

The researchers speculate these results might reflect people’s preference for human explicit advice when it comes to subjective matters such as dating, while people might prefer algorithmic advice on rational political decisions.

Applied to recruitment marketing, this could look like testing to understand if job seekers are influenced to apply based on perceived match or fit using a Job Matching function in a CRM tool, for example. And frequency is more similar to the basic tenet of advertising – seeing an ad multiple times before taking an action – which can be accomplished with programmatic media. As we make our way out of the pandemic and deal with the new challenge of the current limited pool of active job seekers, algorithmic influences on the candidate decision process may become more common. 

But it is a double edged sword; coming at the time when there is greater focus on equitable Artificial Intelligence (AI) and enhanced public education around automation bias – the risk of defaulting to the judgement of AI. 

Read more: https://scitechdaily.com/researchers-warn-ai-algorithms-can-influence-peoples-voting-and-dating-decisions/ [SciTechDaily]

While societal trust in institutions has waned in the current climate, trust in Artificial Intelligence (AI) to be smarter than human instincts or intuition hasn’t abated. The ability to consume more information (data) from which to make decisions, engenders greater confidence in algorithmic guidance…even about ourselves. 

Efforts aimed at reducing the blind-reliance on algorithmic decisions focuses on actively introducing more distrust into systems. If we think about it similar to any other type of learning (conditioning), it will require understanding the point at which the end-user has experienced the expected outcome and decides to trust the system entirely going forward. At that point, a pause is interjected to break the trust pattern. Another attempt at remediation relies on explainable AI, informing the end-user of why the decision is made and some uncertainties inherent to the decision-making process.

A correlation to recruitment would be providing job seekers with a job match score, with information about how the algorithm arrived at the match and possibly additional information that might improve the prediction. Reducing bias may require having more trust in humans to participate in processes that impact their decisions.    

Read more: https://www.technologyreview.com/2021/05/13/1024874/ai-ayanna-howard-trust-robots/ [MIT Technology Review]