Oliver Laas warns of algorithms eroding political legitimacy

Countries are increasingly automating government functions and processes in pursuit of efficiency, but this may come at the cost of political legitimacy, Oliver Laas writes in a Vikerraadio daily commentary published by ERR.

In the Netherlands, an automated system was used to detect welfare fraud. In the United States, similar systems have also been used in attempts to prevent crime. The efforts failed because the algorithms were inaccurate and biased.

Estonia has also considered using data mining to combat money laundering. There has additionally been discussion of using artificial intelligence to identify errors in draft legislation. Krister Kruusmaa has written about how this could influence laws unnoticed.

Laas argues that a greater problem than bias is the lack of transparency. The creators and users of algorithmic systems are often unable to explain their outputs. According to John Danaher, involving such systems in decision-making undermines their legitimacy. He calls this the threat of algocracy.

Danaher defines algocracy as a system in which algorithms are used to collect, organise and transmit the data needed for decision-making within a system of government. By shaping data, algorithms also shape how people in these systems interact with one another and with the wider community.

According to Danaher, algocratic decision-making systems can be integrated into existing legal and bureaucratic systems. They may be based on predictive or data-mining algorithms.

Danaher considers algocracy dangerous because legitimate decision-making processes must involve people and be understandable to them. Growing reliance on algocratic systems limits both public participation and the comprehensibility of decisions, threatening the legitimacy of decision-making.

In Danaher’s view, algocracy is a gradual threat. Efficiency, convenience and accuracy make these systems attractive. But as they are adopted, people gradually become dependent on systems that are not transparent.

Laas writes that the desire for algocracy may not arise solely from the pursuit of efficiency. James C. Scott has argued that states have sought for centuries to make their territories and subjects administratively legible through standardisation. This has included introducing surnames, official languages, standard units of measurement and censuses. Administrative legibility makes state control easier.

Data must also be standardised before it can be processed by algorithms. New types of data may need to be collected to automate decisions. Algocracy can therefore be seen as another step towards making life administratively legible.

Danaher is pessimistic about the future. In his view, society is building a system of government that limits people’s participation. Laas says people should recognise that the price of this development may be political legitimacy.