Sunday, 23 September 2018

Smart nation, unwise citizens

As technology assumes greater dominance, so will our reliance on numbers. But caution is called for as figures are not always reliable.
By Ang Peng Hwa, Published The Straits Times, 22 Sep 2018

Two unrelated incidents recently reminded me how, in our drive to make Singapore a smart nation, we could end up doing some very unwise things.

The first was a comment from a former student of mine who graduated recently with a first-class degree and then applied for a job at a local bank. After taking a psychological analysis test, he was rejected, with the interviewer telling him: You are better off as a musician. That was a revelation to him as he had never given any thought to being a musician.

He had apparently been administered a test by the bank - based on Big Data - and it had determined that his profile would not fit a banking career.

The second was a report earlier this month that a group of five parents was starting a movement to influence Singapore parents to de-emphasise school grades.

The connection between the incidents is this: As we move towards being a smarter nation, numbers will become increasingly important. These numbers may be in the form of a grade in an examination, a score in a personality test, or a rating in an algorithm. And these numbers will increasingly determine our life and our future.

The actions of both the local bank and the group of parents are based on the numbers derived from different sorts of test. But neither blindly relying on a number (as in the case of the local bank) nor selectively de-emphasising it (as with the parents) is quite the correct response.

The bank cannot be faulted for administering some kind of psychological screening test. Large companies, particularly US tech ones, have been using algorithms when hiring staff. The algorithms have proven so successful in predicting the success of hires that once-sceptical managers have been won over. There are accounts of managers who decline to interview applicants but instead say: Get me the applicant with the highest score.

The reason for not wholeheartedly relying on a number is that algorithms, particularly if they truly use artificial intelligence (AI), will make mistakes. That is the nature of true artificial intelligence. Only rules-bound "naive AI" systems do not make mistakes in the narrow sense of being rules-bound. In other words, algorithms can be relied upon most of the time, until they cannot be relied upon. How is one to know? That is the trick.

However, such numbers do give us useful feedback. For example, in jobs that have a high churn, it turns out that one factor that will predict the greater likelihood to remain in the job is distance from work - the nearer the person is to the job, the less likely the churn.

Similarly, grades do give feedback. Yes, grades can be erroneous. But grades in exams, particularly those at the national level such as the PSLE and the GCE, are subjected to multiple checks.

To ignore these grades entirely would be to ignore carefully formulated feedback.

Yes, there is life beyond grades. (At the moment, however, the "life beyond grades" movement seems to be more about "life beyond bad grades" because we know there is a good life beyond good grades.)

The weakness in the Singapore system may be that we place too much emphasis on numbers. We have a tendency to trust numbers, especially when much effort has gone into coming up with them, as in the case of exam grades.

All numbers are subversive because they have the ring of authority, and well-crafted ones even more so as we believe them more. Numbers are particularly appealing and more easily understood by those who do hard sciences. As a result, those who do hard sciences ("easier to score than literature or history") do get better grades and dominate the upper reaches of society. But, like algorithms, grades can be relied upon until they cannot.

We now know, for example, that when children are praised for being intelligent (and how else does one show intelligence but through having a good score), they become more risk averse. There is a downside to good grades.

So how should we treat numbers?

We should acknowledge grades as possible feedback. But we should ponder and question those numbers, and all numbers. Sometimes, perhaps even at critical times, they can be wrong.

In the case of grades, we should make sure that they do indeed capture the weaknesses and strengths of our children. Merely pointing to life beyond bad grades would not help children because it misses the feedback.

For numbers such as those from Big Data or AI, we should question how they are derived and what values and biases may be behind them. There will be a strong tendency to rely on them all the time even when common sense may shout otherwise.

We should not let smart technologies dumb us down.

Ang Peng Hwa is a professor of media law and policy at the Wee Kim Wee School of Communication and Information, Nanyang Technological University. Among his activities is engagement with an artificial intelligence start-up.

No comments:

Post a Comment