DN Debate. Sweden is ready to let the machines make the decisions?

It is feared that a datasystemfel of the employment service has done so up to the seventh decision on the withdrawal of support may have been incorrect. Expressed in figures, this means that up to 70.000 decision may have been wrong. It is, however, no one who knows for sure, it can be both more and fewer, because it is difficult to identify exactly when the system messed up. The employment offices have neither the informed job seekers or employment and if the problem. These erroneous decisions have serious consequences for job-seekers who get their support withdrawn.

the Swedish public administration have begun to replace people with algorithms to decide on everything from income support to child benefit and sickness benefit, or support of the public employment service. Such solutions raises many ethical and integrity related matters. At the same time, as politicians and companies see the automation of the decision-making processes as an important part of the increased digitalisation of society, we are hardly any public discussion about the ethical, legal and societal implications of automated decision-making in public administration. Such a debate is important to ensure that new digital solutions will not undermine the values of equality, justice, and not worsen the situation for already vulnerable people.

that calculates with the help of mathematical models and big data (so called big data) but also by cross-checking the various registers and then leaves the system, a decision if, for example, income support shall be granted or not. Right now there is a little bit of knowledge about what and how it can be automated right now.

at the same time benefit the society of the more efficient ways to make decisions. Citizens may refund payments faster, and everybody is treated in a similar way, in all cases theoretically.

that's Why it was important to contribute to the report ”Automating Society – Taking Stock of Automated Decision-Making in the EU ' , which identifies automation in Europe. The report contains more than 60 examples from 12 countries, as well as an overview of the relevant actors in the process and the current political debates around the automated decision-making. The report shows for the first time, both how the spread of automated decision-making is in Europe, and how varied and non-consistent attempts to address the challenges at the policy level in the different countries.

We are currently in a situation where more automation is in demand both in politics and industry as a shortcut to meet the major societal challenges. At the EU level and within the european Parliament are discussed questions around the automated decision-making with the aim to develop common rules which can control and evaluate the automated system by, among other ethical guidelines. This is very welcome.

at the same time need to pay attention also be directed against the actual application of the automated decision-making at the national level in order to develop a vision of how such a technical, highly complex systems are integrated and used to promote values such as equality and transparency.

the Technology allows for completely new ways to track and predict the behaviors and attitudes based on the large data sets that are collected and merged between the various sectors. When there are consequences, for example, to support, the society must be involved in decisions about how we want to use technical solutions.

One area where automated decision-making has had an impact in Sweden's public administration. Since 2017, the municipalities around the country have started the journey towards automating decision-making, for example, income support. With Trelleborg municipality in the forefront has an algorithm was developed that automatically associates data from the Tax with databases from other government agencies. A decision is generated automatically by the system and delivered to ”the customer”, that is to say, the applicant. The number of administrators has been reduced from eleven to three – thus, a fairly substantial savings. The municipality also claims that they have been able to reduce the number of people with income support significant and thus the costs for the contribution. However, the question is if the number of competent and needy has been reduced or if they were forced out by the new system?

it also shows that It can take a long time to detect the error and to trace the causes.

We live in an everyday life where the decision on the welfare and economy is increasingly delegated to machines, sometimes in contravention of existing legislation or in conflict with the established values of what is right and reasonable. It is therefore important to not only encourage more technological development of machine learning and automation, but also carefully analyse the wider societal impacts and issues that the systems awakens.

at the same time benefit the society of the more efficient ways to make decisions. Citizens may refund payments faster, and everybody is treated in a similar way, in all cases theoretically. Greater efficiency of administration can lead to economic savings but by taking away the personal touch and the way each individual case is also the possibility to take into account the specific circumstances away – at the same time as it is currently very difficult, or impossible, to understand how an algorithm has made a decision. It is precisely this that makes it so problematic with automated decision-making.

We can no longer think of problems related to algorithms and automated decision-making as something that is far away from us in the Nordic countries. We have, therefore, three concrete proposals:

1 There needs to be a broad discussion around the processes and consequences of automated beslutsfattningssystem. the Example of the Trelleborg shows that citizens do not know whether their decisions have been taken by an algorithm or a programme officer. It creates questions about the responsibility of the authority and should receive more attention in the public debate.

2 Citizens should have the opportunity to participate in the actual conversion process of to automated decision-making. Participation has so far been limited unilaterally to the officer but excluded the citizens.

3 the legal regulation of automated decision-making need to be reviewed. In 2018, the adjusted administrative procedure act, but on the municipal level, are fully automated decisions are still excluded. in Addition to this, there are some conflicts between national law and european law. One can interpret it as the change in the administrative procedure act goes against the data protection regulation GDPR's article 22, which explicitly protects personal privacy, and the right to appeal the decision of the authorities.