All recommendations provided are wise and technically rooted, but some passages worth to be highlighted:

Over the past two decades, the ability of machines to challenge and beat humans at complex games has made “quantum” leaps, rhetorically if not in technical computing terms.

In particular, the blanket disclaimer of liabilities attached to virtually all software today should be revisited and revised or rejected if, as it appears, it is inapplicable to many current and likely uses of ADM.

As the defaced stop sign example above demonstrates, minor changes in the perceived world could unpredictably change an appropriate response into an entirely inappropriate one, with possibly lethal consequences

Robust public knowledge of these techniques, without depending predominantly upon industry for research results, is a prerequisite for a broader debate about their acceptability as well as for effective and principled adoption of these techniques by European companies. Improved techniques for explainable automated decision making should be a research priority.

Automated decision making is not just a scientific challenge; it is simultaneously a political, economic, technological, cultural, educational and even philosophical challenge. Because all these aspects are interconnected, it is inappropriate to focus on any one feature of the much larger picture.