Trust in Artificial Intelligence Analysis of the European Commission proposal for a Regulation of Artificial Intelligence

Abstract

Trust in Artificial Intelligence Analysis of the European Commission proposal for a Regulation of Artificial Intelligence Antonio Estella I. Introduction According to the European Commission, one of the main objectives of the regulatory framework that this EU institution is currently proposing in the field of Artificial Intelligence is to "increment trust in the use of artificial intelligence."1 Therefore, this paper explores the issue of trust and AI. The questions that it attempts to answer are the following. Why is trust important? Why is trust important, in particular, in the domain of AI? How does the EU Commission intend to achieve the objective of incrementing trust in the use of AI? Will the proposed regulatory framework achieve its proclaimed end? To answer these questions, this article proceeds as follows. I shall start by reflecting on the importance that trust has for society (section 2). From there, I will define what is to be understood in this paper by trust (section 3). I shall then review the basis of trust (section 4) and shall make a reference to the main sources of evidence on trust (like, surveys and laboratory experiments), and to some of the results that these sources reveal on interpersonal and institutional trust (section 5). In the next section (section 6), I shall go on to analyse specifically the issue of trust in AI, will refer to the existing evidence on the matter, and will review some of the most recent literature on this topic. In the remaining sections (sections 7 and 8), I will describe and analyse the European Commission's proposal for a regulation of AI, and in particular, the part of that proposal that deals with trust in AI. In the last section of this article, I will wrap up the whole argument of this paper and make some conclusions (section 9). The main argument that [End Page 39] will be developed in this paper is that it is inconsequential to speak of trust in AI systems. II. The Importance of Trust Trust has been defined by some authors as the "lubricant of society"2 and by others as "a kind of glue that makes society function."3 Political scientists, economists, and also lawyers have recently centred their intellectual efforts on trying to understand how trust impacts economic growth, development, democracy, justice, and even interpersonal relationships. One particularly clear expression of this renewed interest in trust is the setting up by the OECD of a High Level Group on the measurement of economic performance and social progress.4 The Group started working in 2013. This group convened eight workshops during the years 2014 to 2016. The latest one took place in Paris in June 2016 and was titled: "Measuring Trust and Social Capital." The outcome of this workshop was published in 2018, together with the rest of the reports of the other workshops that have been mentioned, under the title "Trust and Social Capital."5 In this paper, Algan gives ample evidence of how trust is positively correlated with economic growth in general and with economic development in particular.6 The idea is that the more trustworthy a society is, the more it grows and develops in economic terms. The findings of this paper are important since this is the first time that an international institution like the OECD argues that trust should be a necessary component for the measuring of how the nations of the world grow in economic terms.7 [End Page 40] Click for larger view View full resolution Figure 1. Inter-personal trust and income per capita8 On the basis of the previous Figure 1, Algan argues that "countries with higher levels of trust tend to have higher income." For example, Norway has very high levels of trust and has one of the highest incomes per capita of the countries that are included in the previous analysis. An opposite example would be Zimbabwe, with very low levels of interpersonal trust and comparatively low levels of income per capita. Algan acknowledges that there might be problems of reverse causality in analyses on the correlation between trust and economic growth: "one concern has been that this correlation … could go...