History of Mathematics

Day 2 (April 7) @ 13:30–15:00

Gerard Alberts (University of Amsterdam)

Historical introduction: What Is Algorithmic About Algorithmic Knowing?

Algorithms are attributed agency, these days, especially in discussions on deep learning or ArtificiaI Intelligence. In particular they may be said to hold biases; how human a trait.

The pātī- or procedural traditions in mathematics acquired a proper name when a selection of techniques composed by the man from Khwarizm turned into the standard textbook: Al Khwarizmi. While unquestioned reckoning has been a ubiquitous manifestation of mathematical thought in everyday culture, the field of algorithms has gone largely unnoticed by historians of mathematics.

In the  1950s, confronted with the challenge of leaving the computations to such obscure scientific instruments as automatic calculators and still trust the outcomes, Heinz Rutishauser and Hermann Bottenbruch were among the numerical analysts suggesting one should think in terms of algorithms. Alternative answers were suggested, like thinking in terms of control and feedback (under the slogan of cybernetics), or in terms of flowcharts, but “algorithm” won the day as the core concept of computer science. By the end of the decade ALGOL, for ALGOrithmic Language, emerged as the key programming language (with strong participation from Dutch researchers). The algorithmic approach stood for clarity and transparent use of computing machinery, as opposed to the black-boxing culture of cybernetics.

Today, by contrast, the “algorithmic” symbolizes the opaque, the lack of transparency, in which algorithms are vested with mysterious agency; like one would expect in the cybernetic tradition. Time for historians to step in and try to understand the alternating role of “algorithmic”.

Merel Noorman (Tilburg University)

Responsibility for ethical algorithms

Algorithms are not neutral. This has become increasingly clear in recent years with the rising number of scandals involving, for instance, risk-assement, face recognition or predictive analytics algorithms. Algorithms reflect the interests, assumptions and prejudices of those who build them. But it is not just developers that build in their values and worldview; the choices and constraints that underlie the structure of algorithms are shaped within broader networks of human actors, including users, managers, funders, and policy makers,. The ethics of algorithms is thus a responsibility of multiple actors, that each affect these algorithms in different ways. In this talk I will sketch the sociotechnical networks in which algorithms are embedded and discuss how these networks play a role in the development of ethical algorithms. I will focus in particular on the responsibilities of these different actors.

Sennay Ghebreab (University of Amsterdam)

On AI, bias and society

Artificial Intelligence (AI) systems are widely used in society today to make decisions that may have a direct impact on human life, for example in credit risk assessments, employment decisions and criminal suspects predictions. Over the past few years, it has increasingly become clear that human and social biases have made their way into AI systems through data, potentially discriminating entire populations based on gender, race, age and other differences. In this talk I will be addressing two emerging research directions in AI.  The first aims at deeper insight of human and social biases to help identify and mitigate possible harmful biases in AI systems. The second aims at using biased data and AI systems as a magnifying lens to uncover and deal with systemic social bias, inequality and exclusion.