comment 0

Publishing Government Algorithms

On the 1st of February, Malaysians experienced yet another fuel price increase. Which was surprising because the price of oil and the ringgit conversion rate seemed to be favoring a drop. You see in Malaysia, the fuel prices are controlled and subsidized by the government, and it sets the price for petrol at the pump.

In the past, fuel price changes were few and far between, but since 2007, they’ve become a more common occurrence to help the government cope with the erratic pricing of oil. Eventually, a floating price mechanism was introduced, where the government would set the price of fuel on the 1st of every month,¬†and only on the first of each month.

But while citizens knew that the price wouldn’t change throughout a given month, they couldn’t gauge if the price would increase of decrease with each month, because the method of calculating the price was (and still is) a government secret. Which, seems odd, after all, shouldn’t the government share the fuel price calculation to allow citizens some predictability for when fuel prices would increase or decrease?

The deeper question is should government algorithms be secret?

An algorithm is basically a set of steps to be taken to perform a calculation or execute a specific task, if the rules are simple and explicit enough, any idiot could execute any task defined–even a computer. That’s why algorithms form the backbone of our society, every computer, whether its controlling traffic lights at the intersection, or counting free parking lots in the shopping mall, or even predicting the weather–all work on implementing algorithms.

Algorithms control our lives, they decide which Uber Driver you get, who gets audited by the tax department, and which posts appear on your Facebook timeline. Algorithms like the one that powers Google search engine are industrial secrets, tightly held by the companies that own them to maintain their competitive edge. But government algorithms belong to government, and are therefore public property, shouldn’t they be made public?

After all, published algorithms can be vetted to ensure their effectiveness, fairness and accuracy. It gives an assurance to the public, that the algorithm isn’t intentionally sidelining minorities, or favoring privilege groups, and that it actually is effective in carrying out its defined task.

Once we are sure that an algorithm is all good, we can then benchmark the execution of that algorithm, to check that enforcement agencies are working in accordance to a set of a rules and not on the whims and fancies of a little Napoleon. All of which cannot be achieved, if the algorithm driving those decisions are kept under lock and key.

Maybe the government is being lobbied by Oil companies to push the price higher, or maybe aliens are controlling our fuel source, and we’re all here pumping fuel for their amusement. If the government is going to be opaque on the calculations of something as simple as oil prices, it has only itself to blame for the conspiracy theories being sprouted out to explain the unexpected rise of prices in February. A published algorithm solves all of these problems and then some.

Some argue that algorithms need to be secret, in order to guarantee efficacy. For instance if everyone knew how LHDN chose who to audit, tax evaders would do everything in their power to ensure they never met that criteria. But modern algorithms aren’t cast in stone, they are continuously tweaked to adapt to their ever changing world.

For instance, years ago, a few people figured out how to foil Google’s search algorithm, so that all searches for ‘miserable failure’ led to official George W. Bush website. Today the same search takes you to the Wikipedia entry for GoogleBomb, the technique used to pull off this wonderful accomplishment, which no longer works.

That doesn’t mean that Google’s algorithm is perfect, it still fails, but it doesn’t repeat its past failures–human beings may forget what they’ve learnt, but algorithms never forget. Hire enough smart people, and they can help tweak your algorithms from being foiled by tax evaders

In some cases, the government wants things to secret, and hence need the algorithm that drives it to be secret as well. If everyone knew the price of petrol would increase tomorrow, wouldn’t everyone fill their tanks today? And can you imagine the chaos that will ensue?

Well one of the benefits of a published algorithm, is that everyone has access to the same information. Station owners can better prepare for the price increases or decreases, and the public would know well in advance, giving them ample time to fill their tanks–rather than waiting for confirmation at 9pm, and then rushing for it before midnight. In any case, the entire system is used to price changes, and this argument is a red-herring.

Other algorithms, like how Public Universities admit students to their different faculties should also be published for vetting. Being an open and transparent government, isn’t just about publishing the details of decisions but also publishing the steps taken to arrive at those decisions. A published algorithms demonstrate a mature process, one that can be defined, replicated and shared. The flip-side is that if no algorithm is published, we have to assume that the process is immature–undefined at best, irregular and abused at worst.

For thousands of Malaysian students vying for spots in Public Universities, a published algorithm at least gives them a sense of fairness in the system, and would work to correct any misconceptions of bias.

Every makes a wrong decision at some point, but the publishing of the algorithms we use to arrive at those decisions can only be a good thing for the transparency and efficacy of those decisions and (most importantly!) to ensure we don’t repeat those mistakes again.