BUILDING TRUST IN ARTIFICIAL INTELLIGENCE.

• Author: Rossi, Francesca

What is Artificial Intelligence?

Artificial Intelligence (AI) is a scientific discipline aimed at building machines that can perform many tasks that require human intelligence. AI started more than 60 years ago, and includes two main areas of research. One is based on rules, logic, and symbols; it is explainable; and it always finds a correct solution for a given problem, if that problem has been correctly specified. However, it can be used only when all possible scenarios for the problem at hand can be foreseen.

The other area of research is based on examples, data analysis, and correlation. It can be applied in cases where there is an incomplete or ill-defined notion of the problem to be solved. However, this type of AI requires a lot of data, is usually less explainable, and there is always a small margin of error. These two lines of research and ways of thinking about AI are increasingly being combined in order to maximize the advantages of both and to mitigate their drawbacks.

In recent years, many successful applications of AI have been built, mainly because of the convergence of improved algorithms, vast computing power, and massive amounts of data. This provides AI systems with human-level perception capabilities, such as speech-to-text, text understanding, image interpretation, and others, for which machine-learning methods are suitable. These abilities make it possible to deploy AI systems in real-life scenarios that typically have a high degree of uncertainty. Still, current consumer-oriented AI applications where a service is provided to users--from navigation systems to voice-activated "smart" homes--barely scratch the surface of the tremendous opportunity that AI represents for businesses and other institutions.

The main purpose of what can be called enterprise AI is to augment humans' capabilities and to allow humans to make better--that is, more informed and grounded--decisions. At this point, AI and humans have very complementary capabilities, and it is when their capabilities are combined that we find the best results. Typical applications in enterprise AI are decision-support systems for doctors, educators, financial service operators, and a host of other professionals who need to make complex decisions based on lots of data.

A Problem of Trust

It is easy to see that AI will become pervasive in our everyday life. This will certainly bring many benefits in terms of scientific progress, human wellbeing, economic value, and the possibility of exploring solutions to major social and environmental problems. However, such a powerful technology also raises some concerns, such as its ability to make important decisions in a way that humans would perceive as fair, to be aware and aligned to human values that are relevant to the problems being tackled, and the capability to explain its reasoning and decision-making. Since many successful AI techniques rely on huge amounts of data, it is important to know how data are handled by AI systems and by those who produce them.

These concerns are among the obstacles that hold AI back or that cause worry for current AI users, adopters, and policymakers. Issues include the black-box nature of some AI approaches, the possible discriminatory decisions that AI algorithms may make, and the accountability and responsibility when an AI system is involved in an undesirable outcome.

Without answers to these questions, many will not trust AI, and therefore will neither fully adopt...