Outcome Prediction in Finite Systems

Humans live and interact in a world that is becoming more and more complex and diverse. One way to adapt to this changing environment is by establishing proactive control strategies. Proactive control requires anticipating events or effects based on history or knowledge. This is particularly important when managing crisis in areas as disparate as medicine, economy, politics and warfare; when evaluating financial outcomes in economy; or when studying evolutionary patterns in biology.

Outcome prediction involves elements of probability and modeling, is complicated by the complexity (factors/interaction) of the system being studied, and is important in problems of decision-making. Game theory, prisoner's dilemma, and other highly abstract mathematical areas of study are useful in predicting, for example, how cooperative communities evolve and how decisions are made in complex settings. When nonhuman (i.e., nonrational animal behavior) factors are taken into account, however, these theories lose their effectiveness.

Dynamic evolutionary models offer a way to extend the capabilities of game theory and other mathematical models. Underlying dynamic evolutionary models are the notions of reproductive fitness, of stochastic components (based on probability) that account for mutational change, of learning adaptations that stem from past experience, and of intuition. Cladistics, or the systematic classification of organisms based on their evolutionary history, is useful in analyzing a wide variety of evolutionary developments, from molecular complexity to comparative linguistics, and relies on analytical tools such as the tree-of-outcomes approach, which is used to describe the relationships of all descendants to a common ancestor.

Gustavo Caetano-Anollés's research is based on the hypothesis that the tree-of-outcomes approach used in the study of evolutionary biology could also be a useful model in predicting a broad spectrum of outcomes in the domain of human interaction. He intends to predict outcomes by reconstructing past history using a novel strategy that couples cladistics and evolutionary self-learning models based on Bayesian logic, neural networks and Markov-chains. He anticipates that this method will complement game-theoretic and statistical approaches and be used as a general framework for estimating outcomes based on history and learning.

Anollés will conduct studies of macromolecular evolution, reconstructing hierarchical trees depicting the relationship of molecular substructures in noncoding RNA (ncRNA) molecules and describing evolutionary outcomes, allowing him to trace changes along the branches of the tree of outcomes. From this work he hopes to develop a general evolutionary model and demonstrate how it can be used to predict outcomes in non-biological systems.