Thursday, 10 April 2014

How much do we want algorithms to do?

Yesterday I read an engrossing piece by Luke Dormehl in Wired magazine about the prospects for analytic software in the recruitment process. The gist is that algorithms can be a lot more reliable than people in identifying the right candidate, and that it may not be long before organisations rely on technology rather than human judgement to find the right people.

The idea has an immediate appeal in promoting a more genuinely meritocratic workplace. I’m sure that most of us have been frustrated at missing the cut for a job interview when we know we fit the bill. And plenty of us have employed people who had the CV, references and came across great in the interview, but regretted the choice months later. I’ve had experience of both.

So take out those wobbly human judgements and let the algorithms take charge, and we’ll have the right people in the right jobs. Agreed? I suspect there would be more disquiet than enthusiasm.

A lot of managers won’t like it. Getting to choose who work for you is one of the big plus points of being a manager, and it would dent a lot of egos to tell them a computer is more likely to make a good choice. A lot of potential employees would feel dubious, not wanting their worthiness for a job to be assessed by a software programme. There’s a scary element in its implications for our relationship with computers.

It’s also important to remember a question that’s usually asked during recruitment: will they fit in? If you want a machine to answer that you have to combine the data on the candidate with data on their managers, colleagues, and the priorities and dynamics of the company. Will that be readily volunteered? Will it be accurate? The algorithms get more complicated, and become more vulnerable to any distortions and dishonesties. And we all know they come from both sides of the fence.

It’s part of the bigger question of how far cognitive systems can go in replacing humans, something I touched on in a recent white paper for the Chartered Institute for IT. Cognitive computing can do some things better than people, in terms of processing massive quantities of data quicker and more consistently, and can provide important insights at high speed. But they don’t think like people, making the value and moral judgements on which we often rely, and most organisations won’t want to take those judgements out of many of their processes, including recruitment.

I can see more companies using algorithms as part of their process to find new employees; but I bet there won’t be many prepared to drop those face-to-face interviews and take the decisions out of their managers’ hands.

Mark Say is a UK based writer who covers the role of information management and technology in business. See

No comments:

Post a Comment