Postulate Béglé (16.3914): How to bring ethics to algorithms?
Submitted text
The Federal Council is instructed to examine, what can be expected or required of algorithms at home and abroad from an ethical perspective. Algorithms are opaque, their accountability is fuzzy, their duties are limited. How do algorithms work? Who can be contacted in case of misinformation? Are algorithms subject to Swiss law? The ever-growing influence of algorithms must be controlled without diminishing their usefulness.
Justification
Thanks to algorithms, we get hits in our Internet searches. An algorithm is a sequence of instructions used to hierarchize the information consumed by billions of people. The central question is: Which information appears first on our screen? The sequence reflects a value system, even a worldview.
Nowadays, private companies decide about it. The information can be hierarchized according to four criteria: Popularity (number of views), authority (referrals/hyperlinks), reputation (number of retweets/likes), predicted behavior based on traces left on the Internet. However, this hierarchization carries risks.
1. algorithms tend to interfere with the principle of freedom.
In order to engage Internet users, they first display the most accessed information and those that correspond to the user’s opinion, while leaving the rest aside. This creates so-called Filter bubbles. The threat to democracy is therefore all the greater because an increasing number of citizens are Quality media spurn and get information on social networks instead.
2. algorithms exacerbate inequalities.
Algorithms globalize the opinion market and give the “best” excessive visibility. The rest, especially the “average,” is often forgotten: 95 percent of Internet users consume 0.03 percent of the content (according to sociologist Dominique Cardon).
3. this may lead to discrimination.
Dynamic pricing is based on algorithms and can penalize loyal, hurried, or dependent customers by offering them higher prices, and deny attractive offers to customers “without potential.”
With greater transparency and clearer definition of responsibilities, the political, economic, and social power of algorithms could be limited.
Statement of the Federal Council of 25.1.2018
Search, meeting, media, rating and marketplace platforms on the Internet have become an indispensable part of the modern world of life. They are changing the lives of every individual and will increasingly shape the basic values of society. The basis for these systems and platforms are algorithms that enable the necessary data processing. In view of this development, the Federal Council shares the postulate’s assessment that the potential risks posed by algorithms must be investigated if the opportunities are to be exploited sustainably. However, algorithms must not be considered in isolation, but holistically in the context of data processing and the intended functionality of the systems and platforms. Against this background, the Federal Council has taken various measures.
- The measures taken in implementation of the Motion Rechsteiner Paul 13.3841, “Expert Commission on the Future of Data Processing and Data Security,” is investigating the topic of algorithms from various angles. In addition to specific topics – such as digital manipulation (big nudging, filter bubble, etc.) – it is pursuing overarching sets of issues. These include the questions of the extent to which ethical principles in interaction with legal requirements could prevent abuse in data processing in general and in algorithms in particular, and what ethical principles these might be. The report of the expert group with corresponding recommendations is expected in mid-2018.
- In the area of personal data, the ongoing revision of the Data Protection Act various constellations in which personal data is processed by algorithms. Thus, an obligation to inform and hear the data subject is introduced if a decision is made vis-à-vis the data subject that is based exclusively on automated data processing and has legal effects or significant effects on the data subject. With the right to information, the data subject should also be able to demand further information about the result, the occurrence and the effects of the decision. The bill also contains measures on profiling, which is often based on the use of algorithms. Finally, data controllers are to be required to prepare a data protection impact assessment if the processing could lead to a violation of the data subject’s personality or fundamental rights. Within this framework, measures to protect the data subject must also be examined.
- Research and education: In 2015, the Federal Council launched the national research program on Big Data (NRP 75). At the end of last year, various projects dealing with ethical issues in the area of data processing and Big Data were approved as part of Module 2 (societal, regulatory and educational measures).
The measures listed show that the topic of “Algorithms and Ethics on the Net” is already integrated into ongoing activities. For the Federal Council, continuing and strengthening these activities seems the most effective way to address the issue.