marco gherardi

Software as a complex organism

image1

The evolution of software is driven by many interdependent processes, acting during development and maintenance, responsible for solving design-related problems, implementing functionality, responding to user requests, interacting with other programs, and so on. Despite the complex spectrum of forces at play, it is sometimes possible to find consistent regularities, or patterns, pinpointing predictable behavior. These "laws" - akin to physical laws and principles - can suggest to software engineers new metrics and tools to help them control the development process, foresee software properties, and promptly react to abnormal behavior. Moreover, research into the hidden workings of a complex system such as software can reveal, as a treasured byproduct, hints valuable in other contexts, or even a completely new way of thinking at old problems.

Quantitative traits of open collaborations

image1

Not unlike a big city, a large collaborative project evolves in response to the actions of many actors. The evolution of open-source software, for instance, or the redaction of a Wikipedia page, does not react to the indications of a central authority, and is instead the effect of the concerted activity of a possibly large number of users. This new way of collaborating is changing the scale and efficiency of social endeavors, and has catalyzed new ways of thinking about innovation and sustainability. But what are its strengths and weaknesses? What are its consequences on product creation? As huge amounts of data are available from public repositories, the subject is interesting and promising from the standpoint of complexity science. The methods, intuitions and formalism of statistical mechanics are proving powerful allies.

Laws of genome evolution

image1

Genomes carry precious information on evolutionary processes, hiding keys to reveal the regularities, the contingencies, the history, and the fate of organisms and their ecologies. A challenge of theoretical investigations at the interface between physics and biology is the understanding of how universal features emerge in the empirical genomic data. These patterns, sometimes dubbed laws of genome evolution, are linked to phenotypic observables and physiological traits. The toolbox and the vocabulary in this interdisciplinary field encompass more and more concepts from statistical mechanics: random partitioning processes, complex network representations, entropy principles, critical phenomena…

Polymers in 2D and Schramm-Loewner evolution

image1

Polymers, from cellulose and nylon to proteins and DNA, are a long-studied subject, especially from the standpoints of both equilibrium and non-equilibrium statistical mechanics. A simplified description is obtained by considering polymers as random curves, either on the lattice or in continuous space, which brought to the discovery, a few decades ago, of fruitful connections to field theory. A more recent important theoretical advance in the study of the universal behavior of random curves in two dimensions has come with the formulation of stochastic Loewner evolution (SLE), which brings together conformal mapping theory and stochastic processes. The framework provided by SLE is the starting point for explorations and generalizations in many directions (non-equilibrium phenomena, anomalous diffusion, non-extensive statistical mechanics, etc.).