Jessica's research program focuses on coarse-graining and collective computation in nature and their role in the evolution and development of new levels of biological and social organization. She likes to build 'stochastic social circuits' to map the micro to the macro.
Jessica studies a wide range of aggregates, from societies of cells to societies of individuals, seeking common algorithmic principles underlying the emergence of novel, functionally signiﬁcant spatial and temporal scales, and ultimately new kinds of collectives and individuality.
Biological (to include social) systems are composed of heterogeneous components with different information processing capacities and only partly aligned interests, so a natural question is how the components in these systems estimate and control the regularity in their environments to optimally tune their strategies. Jessica is exploring the possibility that components reduce uncertainty by manipulating space and time—producing multi-scale structure—and creating order locally. The increased correlation facilitates collective tuning of the coupling across scales and allows biological systems to manage trade-offs between robustness and adaptability.
Jessica's research has involved development of novel computational techniques (Inductive Game Theory) for extracting strategic decision-making rules from time-series data and using the extracted strategies to construct causal networks or adaptive, stochastic, social circuits that map microscopic dynamics to tunable, functionally important macroscopic states. Using information about biological and cognitive constraints, Jessica and her collaborators reduce the complexity of these circuits (which describe the system’s microscopic behavior) using coarse-graining and dimension reduction techniques to uncover circuit logic and work towards an effective theory for the target macroscopic features.
keywords: computation in nature, complexity, multi-scale, stochastic circuits, robustness, signaling, consensus, coarse-graining, collective behavior and cognition, conflict management, emergence, individuality, information processing, uncertainty reduction, mutual information, criticality, major transitions, causality, social evolution, niche construction
A BIT MORE DETAIL. . .
Biological systems—from cells to tissues to individuals to societies—are hierarchically organized. Think of slime molds, which form when unicellular amoebae come together, mate and grow into plasmoidia—slimy networks of nuclei that can be many meters in size and form reproductive fruiting bodies. Or monkey societies in which slowly changing power structures that constrain behavior are encoded in subordination signaling networks capturing how much consensus there is in the group about who can use force successfully during fights.
Hierarchical organization suggests the nesting of components or individuals into groups, with these groups aggregating into yet larger groups. But this at least superficially privileges space and matter over time and information, and does not apply in all cases, as the power example illustrates. My collaborators and I in the Center for Complexity and Collective Computation are exploring the idea that hierarchical organization at its core is a nesting of functional encodings. We think these functional encodings result from biological systems exploiting space and time to extract information, which in turn facilitates more efficient extraction of energy.
This information hierarchy appears to be a universal property of biological systems and may be the key to one of life’s greatest mysteries—the origins of biological complexity. Our research, which builds upon many years of work at SFI, suggests complexity and the multi-scale structure of biological systems are the predictable outcome of evolutionary dynamics driven by uncertainty minimization.
This recasting of the evolutionary process as an inferential one is based on the premise that organisms and other biological systems can be viewed as hypotheses about the present and future environments they and or their offspring will encounter induced from the history of past environmental states they or their ancestors have experienced. This premise of course only holds if the past is prologue—that is, has regularities, and the regularities can be estimated and even manipulated by biological systems or their components to produce adaptive behavior.
If these premises are correct life at its core is computational and a central question becomes: How do systems and their components estimate and control the regularity in their environments and use these estimates to tune their strategies? We believe the answer to this question, and the explanation for complexity, lies in the manipulation of space and time to create order—low variance—at local scales.
COLLECTIVE COMPUTATION MEANS. . .
Components, whether cells or individuals, extract regularities from their interaction histories with other components and use these regularities to tune their strategies. How components extract these regularities and use them to make predictions is an open question. Preliminary data suggest components can extract regularities by coarse-graining (taking averages over the microscopic behavior), or compressing their interaction histories.
The extent to which a coarse-grained or compressed variable is a good predictor of the fast dynamics depends of course on how much regularity there is at the microscopic level, how much agreement there is among components in their estimates of the average behavior, and also on how quickly changes at the microscopic level can percolate up. A close coupling between scales allows for agile response to change but reduces predictive capacity over longer time frames. On the other hand, timescale separation between the aggregate or macroscopic properties of the system and the microscopic behavior can result in lock-in to a constructed environment that is suboptimal. The challenge for the system is to find the adaptive degree of coupling between microscopic and coarse-grained variables. Systems that have mechanisms for searching over the space of optimal couplings can be said to be computing with multiple timescales.
Of course in social systems at all scales of biological organization there are multiple components interacting and simultaneously coarse-graining. Hence of interest are the collective consequences for social structure of many components searching for regularities and modulating their strategies in response to perceived regularities. Jessica and her colleagues are developing novel computational techniques (Inductive Game Theory) for extracting strategic decision-making rules, game structure, and potentially strategy cost, from time-series data (Dedeo, Krakauer, Flack, 2010; Flack & Krakauer, 2012, Lee et al, in prep). Using the extracted strategies, Jessica and her collaborators construct probabilistic social circuits that map the microscopic behavior to the target macroscopic properties. They then apply biologically principled dimension reduction techniques to the circuit to uncover its logic and build an effective theory for the target macroscopic property. You can read more about this work here.
With two colleagues--David Krakauer and Nihat Ay, Jessica is writing a book on robustness, causal networks, and experimental design that will be published by Princeton University Press.
Jessica Flack is Co-Director of the Center for Complexity and Collective Computation in the Wisconsin Institute for Discovery at the University of Wisconsin, Madison. Jessica received her BA with honors from Cornell University in 1996, studying anthropology, evolutionary theory, and biology. She received her PhD from Emory University in 2004, studying animal behavior, cognitive science, and evolutionary theory. For the next eight years she was in residence at the Santa Fe Institute, first as a Postdoctoral Fellow and then as Research Professor, and finally as Professor. She moved to the University of Wisconsin, Madison in 2011. Jessica’s research has empirical and theoretical components and sits at the interface of evolutionary theory, pattern formation, behavior, cognitive science, computer science, information theory, and statistical mechanics. Although most of her work now is of a computational nature, she has spent thousands of hours collecting large behavioral data sets, including highly resolved time-series, from animal societies, and she conducted the first behavioral knockout study on social systems. In that study, she designed an experiment to disable a critical conflict management function—policing—to quantify its role in social system robustness in an animal society. In addition to peer-reviewed publications, Jessica enjoys writing popular science articles and book reviews. Her work has been covered by other scientists and science journalists in many publications and media outlets, including the BBC, NPR, Nature, Science, The Economist, New Scientist, and Current Biology.
Jessica's nonacademic interests include swimming, surfing, backcountry travel, cooking (chiles and super-spicy food, gnocchi recipes, curries, moles, pastries, sabayon and custards, 'wealthy' apple & fennel pollen pie, pan nero, anything with medjool dates…), gardens and parks, ornamental grasses, conifers (especially those with weeping and irregular forms like Picea pungens weeping blue, Pinus mugo jacobsen, Picea abies cobra, Pinus parviflora tani mano uki, and any Cedrus deodara), tall bearded iris, orchids (esp. Phragmipedium caudatum) art (a diverse bunch, here drawn kind of at random: James Turell, Bruegel, Andrew Wyeth, Eric Fischl, James Drake, Cindy Sherman, Rick Owens, Joseph Cornell, African art, Walton Ford, Aboriginal art, Balthus, Klee, Klimt, Lucien Freud, Odd Nerdrum, Giovanni Bellini, Marcel Dzama, Roberto Matta, Brancusi, etc.. ), all kinds of film (e.g. Alien, Duck You Sucker, Bebette's Feast, The Blind Swordsman: Zatoichi, Seven Samurai, 2001: A Space Odyssey, Blade Runner, The Lives of Others, Terminator, The Royal Tenenbaums, Godzilla and Mothra: The Battle for Earth, Tom Ford's A Single Man, Die Hard, Bottle Rocket, Gosford Park, The Big Gun Down, True Grit--the original and the new one), science fiction, literature (e.g., Tropic of Cancer, Blood Meridian, Suttree, Absalom Absalom!, Sebald's The Rings of Saturn, all of Borges, Sleepless Nights by Elizabeth Hardwick, Ellison's Invisible Man, Gravity's Rainbow, The Recognitions, Lolita, Moby Dick, Carpentier's The Kingdom of this World and Explosion in a Cathedral, Donoso's Obscene Bird of Night, The Mars Trilogy, Tolkien, Hurston's Their Eyes Were Watching God, Baldwin's Tell Me How Long the Train's Been Gone, Beckett's Trilogy), fashion (Rick Owens, futuristic Marni, fully floral Dolce & Gabbana…), and people who are naturally empathic and observant.
A few of her favorite places are Paia and Maui's north shore, Big Sur, maybe parts of East Berlin in the summer, the Storm King Art Center in Mountainville, New York, Telluride, the Weminuche Wilderness, the Grand Canyon of the Tuolumne River, the Grand Tetons, Corsica, Chang Mai, Tanzania, Venice, Morocco, and all of the desert southwest, --particularly Santa Fe, NM, which she considers her home. She lives with David Krakauer, three cats, including one Tonkinese cat, and a dog, who are best buddies. She would have one Tonkinese cat for each harpooner and mate in Moby Dick, but for some reason David does not think this is a good idea. . .
Center for Complexity & Collective Computation
Wisconsin Institute for Discovery
330 N. Orchard Street
Madison, WI 53715