The AI Imperative: Strong Tools Must Have Strong Controls

How AI can transform talent acquisition

In the modern information (and disinformation) saturated world, we tend to think of complex issues as mysterious and difficult to solve. This pertains to our personal lives as well. For instance, do you want to get fit and be the healthiest you can be? Do you want to start a successful company? Or position yourself to retire comfortably at a certain age? 

The reason why many of us are not able to accomplish these things that easily is not because we do not want to, or know how to; rather, it is because the ideal solutions are hard to implement. 

They require studying scientific findings, long-term planning, and expending a frequently large amount of willpower. In short, you must want them badly enough to be willing to make sacrifices to get them.
 

Hiring is the same way – doing it right is not mysterious. In fact, I would argue it has basically been solved. However, unfortunately, there are no silver bullets, so you can forget about the idea of one shiny new AI tool fixing everything. 

And while there remains a lot of irrational exuberance over new AI-based technologies, the truth is more nuanced: AI has an important place in the ideal hiring process but must be carefully managed to prevent any potentially harmful consequences. 

What exactly is “artificial intelligence” (AI) anyway? According to the computer scientist Astro Teller, “AI is the science of how to get machines to do the things they do in movies.” 

A less irreverent approach is to think of AI simply as advanced data analysis techniques. In contrast to traditional statistical analysis, which is great at interpreting clean numerical data like test scores, AI (especially deep learning) is extremely good at making sense of messy, non-numerical, unstructured information. 

This is why we now have cars that can process unstructured, real-world data from the environment to drive themselves and how software can interpret radiological scans as well as, or more accurately, than trained physicians. 

How AI Can Transform Talent Acquisition

In hiring, messy data abounds – resumes, social media, video, free-text, open-ended interview responses, and more. Until just a few years ago, algorithms could not do much with such input. But new AI tools are increasingly able to make sense of such data.

While AI should not be viewed as a cure-all, it can improve many aspects of the hiring process, including:
 

• Scoring interviews – Interviews are part of just about every hiring process on earth. Done poorly, they can lead to bad and biased hiring decisions. But scientific, structured, behavioral-based interviews can be among the most valid and fair ways to make a hiring selection. 

AI can now make interviews even easier and more objective by automatically scoring interview responses against job-relevant competencies. 
 

• Better analytics – Traditionally, hiring processes have multiple, discrete steps like interviews, assessments, background checks, and more. The hiring manager is left to summarize all that information to make the best possible decision. But the industry is making more progress every day in having algorithms do all that for us. 

AI can quantify data from each step (as with scored interviews) and then a process called multiple objective optimization can combine that data in a way that is optimized for predictive power (or validity) and fairness.  
 

• Candidate communications – Many applications of AI have been invented to attempt to communicate with candidates in a more timely and helpful manner, including chatbots. Done well, such technology can help candidates feel more informed about the employer’s process and their status in it. 

 

• Shorter processes – No one likes long, cumbersome hiring processes and AI can increasingly help us shorten the experience, while simultaneously ensuring the results are even more predictive of new hire success.

Research shows candidates do not necessarily prefer a short hiring process – they will tolerate a longer one if it is engaging. Candidates are driven away by frustrating, burdensome, disconnected processes. You don’t lose candidates during an interaction; you lose them between interactions. 

Why You Shouldn’t Get High on Your Own AI

If AI-based tools are not thoughtfully designed or implemented, potential downsides are inevitable and numerous. Know what to look for:
 

• Bias – AI has a reputation for perpetuating bias – you’ve surely seen the headlines. Indeed, all goes wrong when algorithms trained on biased data are then scaled up as part of automated decision-making processes, effectively magnifying that bias over a huge population. To prevent this kind of proliferation, developers must be thoughtful about the data they use to build algorithms. Even then, the results of algorithmic tools must be monitored continually to ensure they are not causing unanticipated outcomes. 

 

• Hype – The term “artificial intelligence” makes marketers giddy with visions of increased revenue.  But buyers must rigorously evaluate what a product can do for them. Ask how the tool was built, how much and what type of data was used to train it, how it is continuously evaluated, and what outcome measures it has been proven to improve. Ultimately, simply having an AI component in your hiring process doesn’t mean much – it is if and how it improves the process that matters. 

 

• Dehumanization – Algocratic orchestration is when human work is directed by algorithms. And our working lives increasingly will be governed by algorithms. In hiring, overuse of automation can leave candidates feeling cold, especially when they find the limits of the algorithms’ knowledge. For example, when that chatbot clearly doesn’t understand what your candidate is asking, your candidate is going to be frustrated and not impressed with your hiring technology. And if you have tried to frame that bot as human, it will feel false and deceitful. Automation should only be used to enhance the candidate experience, not automate the actual, vitally important human component.

Ultimately, any tool, AI-based or not, needs to support at least four important hiring outcomes: efficiency, effectiveness, candidate engagement, and fair decision support. Tools that only support one may not add anything useful to your hiring process – or worse, do so at the expense of another outcome. 

Again, there are no silver bullets. The best hiring process balances smart tools with achieving the best all-around results. 

The good news is, AI and other modern tools are well positioned to do exactly this.

Eric Sydell is EVP and I/O psychologist at Modern Hire. Visit Modernhire  Connect Eric Sydell