Bayesian belief propagation is a powerful concept used in machine learning and the social sciences to understand complex problems involving probabilistic inference in groups. It is particularly useful in heterogeneous networks, where the nodes and edges represent different types or characteristics. This article explains how Bayesian belief propagation works, and discusses its role in the emergence of filter bubbles.

In the words of internet pioneer Marc Andreesen, software will inevitably eat the world. “The world" in this case is the world of physical products, and “software" is any kind of digital representation that can emulate or replace this product. A very prominent example of this prediction is the replacement…

Thinking about decentralized systems means thinking about how an organization can agree upon what it holds to be true and how this relates to the actual ground truth. — Centralized “truth” A “single source of truth” is a concept from information science, but it’s so universal that it can be applied to a whole variety of scenarios and organizations. It means that whenever participants disagree about the state of the world, or a particular aspect of it, a designated participant gets…

“Decentralization" is a popular term used to bracket a set of interrelated technologies including blockchain, crypto tokens, NFTs. It’s also a fuzzy term that has been used in subtly different ways in different disciplines throughout history, so the question “what is a decentralized system", or, more to the point, “is…

A few notes on coordination, adaptation, and the role of institutions. — “Both the organization theorist Chester Barnard and the economist Friedrich Hayek took adaptation to be the main purpose of economic organization, but with differences.” — Oliver Williamson Nobel Lecture Coordination has become an increasingly important C-word in business and economics, based on the emerging realization that not all economic activity…

In 1991, social scientist and Nobel laureate Herbert Simon invited us to participate in a somewhat alien-sounding thought experiment. Imagine an extraterrestrial observing human behavior with a special telescope. Instead of showing us as we are, the telescope allows the observer to recognize social structures: organizations appear as green blobs, interactions between organizations as a network of red lines between the blobs.

In 1956, the fledgling Graduate School of Industrial Administration (GSIA) at Carnegie Institute of Technology received an IBM 650 computer to use jointly with the engineering and the mathematics departments. In the coming years, this acquisition led to a number of research activities which not only produced multiple economics Nobel Prizes and Turing Awards, but also two very distinct, somewhat incompatible economic frameworks: bounded rationality and rational expectations.

One of the intriguing aspects of venture building is the contrast between two very different types of engagement: the creation of new potential opportunities and the cutting off of those opportunities that, after some probing, don’t appear viable. — The two steps of venture building These two types require two very different modes of work. In operations research they are known as “branch and bound”. Within an established company, “research” and “development” are typically two separate organizations within overall product R&D.

I was recently asked what the most obvious long-term effect of the “whiplash of 2020” was. My answer was that every social change we’d usually expect to unfold within twenty years would suddenly only take five. — In the famous technology adoption curve popularized by Geoffrey Moore, new ideas often linger for ten or more years in the segments reserved for innovators and early adopters: people enamored with technology and willing to put up with its drawbacks.