By Anne Kaun and Julia Velkova

The Swedish government aims to place the country at the forefront of the technological development around Artificial Intelligence (AI) and automated decision-making (ADM). In order to accomplish this, it kicked off several strategic initiatives since 2017, including extensive government reports on self-driving cars and a commission for Artificial Intelligence. It also provided broad funding for the improvement of knowledge and skills on Artificial Intelligence for universities and university colleges. Besides the Swedish government, there are a number of additional stakeholders who are investing in knowledge production and the development of AI on a large scale. One initiative that should be mentioned here is the Wallenberg Autonomous Systems and Software Programme (WASP). It granted €100 million to two leading universities in Sweden to develop machine learning, AI, and the mathematical apparatus behind them over the next eight years. At the same time, automated decision-making is increasingly implemented in public service institutions, especially on the municipality level. However, regulation lags behind this development and the public is largely unaware of automated decision-making in the public sector and welfare institutions. The private sector already offers a range of different ADM-based solutions, ranging from office workflow optimisation, credit scoring and juridical support to self-driving lorries and the automatic detection of dyslexia among school children.

Political debates on aspects of automation – Government and Parliament

Swedish AI strategy

In February 2018, the Swedish parliament published a strategy for the national adoption of AI. [SE 1] It is framed as complementing the already existing digitisation strategy. The AI strategy argues that Swedish competitiveness and welfare depend on the development of AI and digitisation. It suggests that AI needs to be implemented in a ‘sustainable’ way, which rests on designing applications that are “ethical, secure, reliable and transparent”. The strategy identifies as a crucial prerequisite the need for more expertise to develop and use AI in different parts of society including companies, municipalities, regional administrations and governmental agencies. It is, however, debatable to what extent different groups in society are actually involved. At present, most of the funding and strategic development takes place in the universities and as support for business environments. The need for expertise is linked to the need for educational institutions to produce experts, particularly engineers. [SE 2] Cybersecurity and collaboration with the defence sector are also outlined as important, as well as building EU-wide partnerships around AI. Engineers are singled out and given priority, even though the strategy admits that the field requires the involvement of other professions.

The strategy stresses that automated decision-making by governmental agencies poses moral dilemmas and concerns. The strategy argues that it is necessary to work on AI at a pan-EU level and desirable to achieve both a dialogue and a leading role in setting EU-wide standards, best practices and infrastructure for implementing AI. [SE 3]

New governmental committee for coordinated and accelerated development of policies related to the fourth industrial revolution (2019-2020)

In August 2018, the Swedish government created a committee in charge of developing policies and accelerating automation across industries in Sweden. [SE 4] The government justified the creation of the committee with the accelerated adoption of “the fourth industrial revolution”, which it described as characterised by “constant connectivity, smaller and more powerful sensors, artificial intelligence and machine learning”. A major task for the group is to suggest policy developments. The committee references the governmental report on autonomous vehicles as an example that articulates important challenges to legislation, i.e. defining rules for not-yet-realised technologies and their areas of application. The committee is expected to support the work of the government by identifying policy challenges, contributing to the reduction of legal uncertainties and to speed up policy development, in particular in the following areas: precision medicine, connected industries (this refers to ‘smart’ factories or industrial manufacturing that has been digitised), connected and automated vehicles, transport and systems. The committee is tasked with producing analyses of barriers for the adoption of “the fourth industrial revolution”, such as existing regulatory frameworks, to map the need for adjusting existing regulatory frameworks, to continuously come up with suggestions for the government regarding policy developments, promote a dialogue between relevant governmental agencies and regional actors, educational institutions, the non-governmental sector, and business for efficient collaboration concerning policy-developments. It is, however, not stated how, and more precisely which of these actors will be involved. In addition, the committee should seek collaboration with international actors such as EU institutions, the OECD, the World Economic Forum and other countries and international institutions. A referee group with representatives from the state and governmental agencies, regional actors, the business sector and organisations with experience in policy development will be assigned to support the committee. The committee is supposed to produce a mid-term report for the years 2019 and 2020 and a final report is due on December 31, 2021. The committee started its work at the beginning of October 1, 2018. The Head of the committee is Jon Simonsson, [SE 5] a former entrepreneur and CEO, whose prior work for the government included being head of the section for innovation within the Ministry of Enterprise and Innovation.

Governmental report on policy development related to ­autonomous driving vehicles

In March 2018, a governmental committee presented a 1,314-page report [SE 6] that evaluates the legal framework for introducing automated vehicles into ordinary traffic and makes concrete policy suggestions to improve it. The task for the committee was to consider a fast introduction of vehicles with automated functions as part of a radical transformation of the transport sector. In the English summary of the report, the authors state its main outcomes as follows: “In the opinion of the committee, multi-stage development of regulations is required to deal with developments in the field of automated, electrified and digitised mobility so that this development can take place in a safe, sustainable manner. The committee’s proposals are intended to commence adaptation of the regulations so that these do not impede the development of new solutions for enhanced attainment of transport policy targets. One difficulty regarding this work has involved developing a regulation for a phenomenon that is not yet available on the market, namely fully automated vehicles capable of replacing the driver. The committee has attempted to suggest solutions that provide enhanced opportunities for testing and introducing advanced automated functions in vehicles in the short term, as well as certain fully automated vehicles. However, these solutions can primarily be used even when a broader introduction becomes possible.” [SE 7] The suggested solutions include both changes in the current regulations (for example in the regulation of penalties in traffic and for drivers’ licences), as well as proposals for completely new regulations and laws (for example a new law and regulation on self-driving cars).

Political debates on aspects of automation – Civil Society and Academia

addAI initiative – a policy initiative

The addAI initiative [SE 8] is a collaboration between experts in academia, government and companies to discuss and explore the impact of smart algorithms and AI on society. Members of the initiative have organised workshops and speak at public events such as the Swedish Internet Days (every year in November).

The initiative works on questions like:

The Internet Foundation

The Internetstiftelsen (Internet Foundation in Sweden, IIS) [SE 9] is an independent and public organisation that works—in their own words—towards a positive development of the Internet. The foundation is based in Sweden and is responsible for the country code top level domain .se as well as .nu. Besides supporting technological innovation and research about the Internet, the foundation publishes guides and teaching materials to enhance digital public education as well as knowledge production. In relation to ADM, the foundation has produced several publications on digital philosophy and algorithms for the general public.

Swedish Trade Union Confederation

The Landsorganisationen i Sverige (Swedish Trade Union Confederation, LO) published a report [SE 10] about how work has changed over the past 20 years and concludes that it has been increasingly impoverished. Impoverished work refers to work that requires less expertise, shorter training time, work that involves less variation in work operations, but rather involves repetitive tasks, work that does not require a lot of learning and where the employee has less options to work creatively. The report’s main conclusion is that most workplaces are characterised by digital taylorism where assignments are increasingly standardised. This, of course, has consequences if standardised assignments are increasingly automated. The report refers to a number of positive examples where automation has led to the upskilling of workers and larger variations in work assignments.

The Union

Unionen (The Union) is a white-collar union that has conducted opinion polls among its members to map current applications of automation in the service sector. Based on the poll the union will develop a strategy for how to approach ADM and automation more generally. [SE 11] The main questions of the poll are related to the number of jobs that have disappeared because of automation and how companies are enhancing the competitiveness of their employees in that context.

Wallenberg Autonomous Systems and Software Program

A private fund and initiative called the Wallenberg Autonomous Systems and Software Program (WASP) [SE 12] has granted €100 million [SE 13] to two leading universities in Sweden to develop machine learning, AI, and the mathematical apparatus behind them in the next 8 years. 300 doctoral students will be trained, and an additional €300 million will be added to the main funding of the program.

Regulatory and self-regulatory Measures

Legislative changes of the Administrative Act

A legislative change that entered into force in July 2018 allows governmental agencies to implement automated decisions. [SE 14] The law regulating the procedures of Swedish governmental institutions (Förvaltningslagen) was amended with a new clause under the headline “How can a decision be taken?” It states that “A decision can be taken by a decision-maker responsible alone, or together with others, or in an automated manner”. According to the commentary about the law, an automated decision means a decision made by machines without a person in the public authority taking any active part in the decision-making process. The decision builds upon propositions made in 2014 by a former Swedish committee for e-governance. As the law might be interpreted as being inconsistent with article 22 of the GDPR, an explicit commentary says that the specifics of the law include sufficient protection of individuals and the right to appeal.

Even though this change was implemented, the law does not cover the legislative framework that governs municipalities. This framework explicitly prohibits automated decisions-systems can only generate suggestions for decisions, but a person needs to approve them and assume responsibility. Indeed, the law makes an explicit distinction between automated decision-making and automated decision support. Where and when to draw the line between both is a topic of debate in Sweden at the moment. Municipal officials say that a change in the law to allow automated decision-making is ‘desirable’. The law requires that a detailed motivation for the reasoning that led to a decision should be attached. A Swedish expert in municipal law suggests [SE 15] that the requirement for motivation sets limits on automation which takes decisions with little consideration of personal circumstances. However, he adds that future systems could be made to cater for such details. In the meantime, the current legislative framework [SE 16] makes illegal solutions of automated welfare decisions, such as those implemented in the municipality of Trelleborg, and multiplied in other cities in Sweden (see the ADM in Action section below). The lack of such clarity makes the forced relocation of employees, who used to work in the processing of welfare decisions, to other parts of the municipal authority potentially problematic.

ADM in Action

Airhelp – Automation of juridical support

Since November 2017, Airhelp employs an advanced algorithm called Lara to analyse compensation claims from airlines and evaluate in real-time how likely it is that a passenger will be compensated for interference to flight and travel plans. [SE 17] The algorithm bases the decision on a number of variables including flight status, flight statistics and weather forecast. In that way, the algorithm has automated parts of the responsibility in terms of arguing for compensation claims and decision-making.

Credit scoring

Lendo.se [SE 18] provides an automated calculation of individual risk with 25 different banks and loan-givers in Sweden based on their unique risk-assessment criteria before taking out a loan. It then issues an automatically generated document that states whether the applicant should be given a loan or not. The document issued by Lendo is valid for 30 days. Similar services are offered by UC.se [SE 19] and Bisnode.se [SE 20], among others, who automate risk assessment, but also offer other services like automated marketing based on algorithmic audience profiling that takes digital footprints as an input. [SE 21]

Einreid – Automated lorries

The Swedish startup Einreid [SE 22] pioneers “intelligent movement” through a cloud-based, real-time computing truck. The truck decides on its driving route in real-time using “big data analysis” that “integrates customer data, traffic data, etc. to optimise delivery time, battery life, and energy consumption—making the journey from A to B as efficient as possible”—according to the company’s website.

Hedvig – Automated home insurance

The start-up company Hedvig [SE 23] [SE 24] [SE 25] has gotten a lot of publicity in the media for having automated the filing of insurance claims. The app uses a voice input and after that it automatically writes the claim, sends it, and the insurance company automatically processes it and disburses the payment. At the moment, only home insurance is automated in this way.

Lexplore – Automatically detecting dyslexia in children

Lexplore [SE 26] is employing AI to read and analyse the eye movement of children reading from a screen to detect dyslexia. In Sweden, the service is directly sold to both schools and municipalities. Lexplore is currently being piloted in different states of the US.

Social benefits – The Trelleborg model

Since 2017, Trelleborg has automated parts of its decision-making when it comes to social benefits. [SE 27] New applications are automatically checked and cross-checked with other related databases (e.g. the tax agency and unit for housing support). A decision is automatically issued by the system. The number of caseworkers has been reduced from 11 to 3 and the municipality argues that they have considerably reduced the number of people receiving social benefits. They have been heading a pilot project to export their automation model to 14 additional municipalities and have received several innovation prizes. However, applicants and citizens have not been explicitly informed about the automation process. During the implementation process in another municipality, more than half of the caseworkers left their jobs in protest. [SE 28]

Anne Kaun

Anne Kaun is Associate Professor at the Department for Media and Communication Studies at Södertörn University, Stockholm. Her research combines archival research with interviews and participant observation to better understand changes in how activists have used media technologies and how technologies shape activism in terms of temporality and space. In her most recent book, “Crisis and Critique”, she explores the role of media in the shaping of social movements and resistance to capitalism. Furthermore, she is interested in different forms of digital and algorithmic activism and is studying the consequences of automation in public service institutions. She also explores prison media, tracing the media practices and media work of prisoners since the inception of the modern prison system. Anne serves as a chair of the Communication and Democracy section of ECREA and is vice-chair of the Activism, Communication and Social Justice Interest Group within ICA.

Julia Velkova

Julia Velkova is a digital media scholar and post-doctoral researcher at the Consumer Society Research Centre at the University of Helsinki. Her research lies at the crossing between infrastructure studies, science and technology studies and cultural studies of new and digital media. She currently works on a project which explores the waste economies behind the production of ‘the cloud’ with focus on the residual heat, temporalities and spaces that are created in the process of data centres being established in the Nordic countries. Other themes that she currently works with include algorithmic and metric cultures. She is also the Vice-Chair of the section “Media Industries and Cultural Production” of the European Communication and Research Education Association (ECREA). Her work has been published in journals such as New Media & Society, Big Data & Society, and International Journal of Cultural Studies, among others.