Budaev, S., Kristiansen, T. S., Giske, J., & Eliassen, S. (2020). Computational animal welfare: Towards cognitive architecture models of animal sentience, emotion and wellbeing. Royal Society Open Science, 7, 201886. doi:10.1098/rsos.201886.
Abstract. To understand animal wellbeing we need to consider subjective phenomena and sentience. This is challenging, since these properties are private and cannot be observed directly. Certain motivations, emotions and related internal states can be inferred in animals through experiments that involve choice, learning, generalisation and decision-making. Yet, even though there is a significant progress in elucidating the neurobiology of human consciousness, animal consciousness is still a mystery. We propose that computational animal welfare science emerges at the intersection of animal behaviour, welfare, and computational cognition. By using ideas from cognitive science, we develop a functional and generic definition of subjective phenomena as any process or state of the organism that exist from the first-person perspective and cannot be isolated from the animal subject. We then outline a general cognitive architecture to model simple forms of subjective processes and sentience. This includes evolutionary adaptation which contains top-down attention modulation, predictive processing and subjective simulation by reentrant (recursive) computations. Thereafter we show how this approach utilises major characteristics of the subjective experience: elementary self-awareness, global workspace and qualia with unity and continuity. This provides a formal framework for process-based modelling of animal needs, subjective states, sentience and wellbeing.
Weidner, J., Jensen, C. H., Giske, J., Eliassen, S., & Jørgensen, C. (2020). Hormones as adaptive control systems in juvenile fish. Biology Open, 9(2), bio046144. doi:10.1242/bio.046144.
Abstract. Growth is an important theme in biology. Physiologists often relate growth rates to hormonal control of essential processes. Ecologists often study growth as a function of gradients or combinations of environmental factors. Fewer studies have investigated the combined effects of environmental and hormonal control on growth. Here, we present an evolutionary optimization model of fish growth that combines internal regulation of growth by hormone levels with the external influence of food availability and predation risk. The model finds a dynamic hormone profile that optimizes fish growth and survival up to 30 cm, and we use the probability of reaching this milestone as a proxy for fitness. The complex web of interrelated hormones and other signalling molecules is simplified to three functions represented by growth hormone, thyroid hormone and orexin. By studying a range from poor to rich environments, we find that the level of food availability in the environment results in different evolutionarily optimal strategies of hormone levels. With more food available, higher levels of hormones are optimal, resulting in higher food intake, standard metabolism and growth. By using this fitness- based approach we also find a consequence of evolutionary optimization of survival on optimal hormone use. Where foraging is risky, the thyroid hormone can be used strategically to increase metabolic potential and the chance of escaping from predators. By comparing model results to empirical observations, many mechanisms can be recognized, for instance a change in pace-of- life due to resource availability, and reduced emphasis on reserves in more stable environments.
Jensen, C. H., Weidner, J., Giske, J., Budaev, S., Jørgensen, C., & Eliassen, S. (2020). Hormonal adjustments to future expectations impact growth and survival in juvenile fish. Oikos, oik.07483. doi:10.1111/oik.07483
Abstract. Evolutionary ecology often studies how environmental factors define optimal pheno- types without considering the bodily mechanisms involved in their regulation. Here we used a dynamic optimisation model to investigate optimally concerted hormonal control of the phenotype. We studied a semi-realistic situation where hormonal con- trol of appetite, metabolism and growth acts to prepare juvenile fish for an uncertain future with regard to food availability. We found a bottom–up effect in that hormone levels varied across environments and affected a range of phenotypic changes. We also describe a top–down effect as natural selection varied across environments, which affected evolutionary optimisation of hormone levels. These combined top–down and bottom–up effects produced a hormone-regulated phenotype that adjusted its forag- ing intensity and risk-taking in adaptive ways depending on the differences between current and expected long-term environmental conditions. Hence, understanding the response of these fish to their current conditions also requires an understanding of their future expectations. We found that when food availability was low, it was optimal for the juvenile fish to have low growth hormone, thyroid hormone and orexin levels, contrary to when food availability was high when these levels were higher. Individual variation emerged from the individually experienced food availability trajectories: Those that on average experienced higher food availability grew faster and had higher short-term mortality risk. They also had higher survival probability throughout the growth period. The opposite was true for individuals experiencing lower food availabil- ity. Hormonal mechanisms that often are overlooked by ecologists are thus important in the ultimate adaptive control of both behaviour and physiology, thereby impacting fitness through growth and survival.
Budaev, S., Jørgensen, C., Mangel, M., Eliassen, S., Giske, J. (2019). Decision-making from the animal perspective: Bridging ecology and subjective cognition. Frontiers in Ecology and Evolution, 7, 164, doi:10.3389/fevo.2019.00164.
Abstract. Organisms have evolved to trade priorities across various needs, such as growth, survival, and reproduction. In naturally complex environments this incurs high computational costs. Models exist for several types of decisions, e.g., optimal foraging or life history theory. However, most models ignore proximate complexities and infer simple rules specific to each context. They try to deduce what the organism must do, but do not provide a mechanistic explanation of how it implements decisions. We posit that the underlying cognitive machinery cannot be ignored. From the point of view of the animal, the fundamental problems are what are the best contexts to choose and which stimuli require a response to achieve a specific goal (e.g., homeostasis, survival, reproduction). This requires a cognitive machinery enabling the organism to make predictions about the future and behave autonomously. Our simulation framework includes three essential aspects: (a) the focus on the autonomous individual, (b) the need to limit and integrate information from the environment, and (c) the importance of goal-directed rather than purely stimulus-driven cognitive and behavioral control. The resulting models integrate cognition, decision-making, and behavior in the whole phenotype that may include the genome, physiology, hormonal system, perception, emotions, motivation, and cognition. We conclude that the fundamental state is the global organismic state that includes both physiology and the animal's subjective “mind”. The approach provides an avenue for evolutionary understanding of subjective phenomena and self-awareness as evolved mechanisms for adaptive decision-making in natural environments.
Budaev, S., Giske, J. & Eliassen, S. (2018) AHA: A general cognitive architecture for Darwinian agents. Biologically Inspired Cognitive Architectutes, 25, 51-57. doi:10.1016/j.bica.2018.07.009 [supplementary pdf]
Abstract. Unified theories of cognition have traditionally played a vital role in understanding the human mind. In the animal behavior field, however, acceptance of holistic views on the behavioral phenotype that includes diverse cognitive and behavioral traits is rather slow. Studying adaptation and evolution of behavior, especially complex cognition and decision making, requires integrative models applicable to a range of species. We describe a general cognitive architecture and a modeling framework for studying evolution and adaptation of behavior and cognition that we call Adapted Heuristics and Architecture (AHA). AHA is non-symbolic, rule-based and grounded in general neurobiological mechanisms. It integrates the whole organism with its genome, physiology, hormones, perception, emotion, motivation and cognition in an agent based model environment. The method lets us investigate various scenarios for the evolution of cognition, decision making and emergence of subjective phenomena. We illustrate the potential feasibility of the framework with a model of simple forms of self-awareness.
Eliassen, S., Andersen, B.S.B.S., Jørgensen, C., Giske, J., Jorgensen, C. and Giske, J. (2016) From sensing to emergent adaptations: Modelling the proximate architecture for decision-making. Ecological Modelling 326, 90–100. doi:10.1016/j.ecolmodel.2015.09.001
Abstract. During the past 50 years, evolutionary theory for animal behaviour has branched into different methodological frameworks focussing on age-, state-, density-, and frequency-dependent processes. These approaches have led to valuable insights in optimal responses, state dependent choices, and behavioural strategies in social contexts. We argue that time is ripe for an integration of these methodologies based on a rigorous implementation of proximate mechanisms. We describe such a modelling framework that is based on the architectural structures of sensing and information processing, physiological and neurological states, and behavioural control in animals. An individual-based model of this decision architecture is embedded in a genetic algorithm that finds evolutionary adaptations. This proximate architecture framework can be utilized for modelling behavioural challenges in complex environments, for example how animals make behavioural decisions based on multiple sources of information, or adapt to changing environments. The framework represents the evolution of the proximate mechanisms that underlie animal decision making, and it aligns with individual-based ecology by emphasizing the role of local information, perception, and individual behaviour.
Andersen, B.S., Jørgensen, C., Eliassen, S. and Giske, J. (2015) The proximate architecture for decision-making in fish. Fish and Fisheries, 680–695. doi:10.1111/faf.12139
Abstract. Evolution has since the very beginning resulted in organisms which can sort fitness-related information from noise, evaluate it and respond to it. In animals, the architecture for proximate control of behaviour and physiology has been gradually evolving since before the Cambrian explosion of animal phyla. It integrates many different survival circuits, for example for danger, feeding and reproduction, and operates through reflexes, instincts, homeostatic drives and precursors to human emotions. Although teleost brains differ substantially from the much better understood brains of terrestrial vertebrates, their anatomy, physiology and neurochemistry all point towards a common and malleable architectural template with strong and flexible effects on fish behaviour and elements of personality. We describe the main components of this architecture and its role in fish behaviour from the perspectives of adaptation, evolutionary history and gene pools. Much research is needed, as several of the basic assumptions for architectural control of behaviour and physiology in teleosts are not thoroughly investigated. We think the architecture for behavioural control can be used to change ecosystem models from a bottom-up perspective to also include behaviourally mediated trophic cascades and trait-mediated indirect effects. We also discuss the utility of modelling based on proximate architectural control for fish welfare studies.
Giske, J., Eliassen, S., Fiksen, O., Jakobsen, P.J., Aksnes, D.L., Mangel, M. and Jorgensen, C. (2014) The emotion system promotes diversity and evolvability. Proceedings of the Royal Society B: Biological Sciences 281, 20141096. doi:10.1098/rspb.2014.1096 [Fortran code for H18 doi:10.5061/dryad.m6k1r]
Abstract. Studies on the relationship between the optimal phenotype and its environment have had limited focus on genotype-to-phenotype pathways and their evolutionary consequences. Here, we study how multi-layered trait architecture and its associated constraints prescribe diversity. Using an idealized model of the emotion system in fish, we find that trait architecture yields genetic and phenotypic diversity even in absence of frequency-dependent selection or environmental variation. That is, for a given environment, phenotype frequency distributions are predictable while gene pools are not. The conservation of phenotypic traits among these genetically different populations is due to the multi-layered trait architecture, in which one adaptation at a higher architectural level can be achieved by several different adaptations at a lower level. Our results emphasize the role of convergent evolution and the organismal level of selection. While trait architecture makes individuals more constrained than what has been assumed in optimization theory, the resulting populations are genetically more diverse and adaptable. The emotion system in animals may thus have evolved by natural selection because it simultaneously enhances three important functions, the behavioural robustness of individuals, the evolvability of gene pools and the rate of evolutionary innovation at several architectural levels.
Giske, J., Eliassen, S., Fiksen, Ø., Jakobsen, P.J., Aksnes, D.L., Jørgensen, C. and Mangel, M. (2013) Effects of the emotion system on adaptive behavior. The American naturalist 182, 689–703. doi:10.1086/673533
Abstract. A central simplifying assumption in evolutionary behavioral ecology has been that optimal behavior is unaffected by genetic or proximate constraints. Observations and experiments show otherwise, so that attention to decision architecture and mechanisms is needed. In psychology, the proximate constraints on decision making and the processes from perception to behavior are collectively described as the emotion system. We specify a model of the emotion system in fish that includes sensory input, neuronal computation, developmental modulation, and a global organismic state and restricts attention during decision making for behavioral outcomes. The model further includes food competition, safety in numbers, and a fluctuating environment. We find that emergent strategies in evolved populations include common emotional appraisal of sensory input related to fear and hunger and also include frequency-dependent rules for behavioral responses. Focused attention is at times more important than spatial behavior for growth and survival. Spatial segregation of the population is driven by personality differences. By coupling proximate and immediate influences on behavior with ultimate fitness consequences through the emotion system, this approach contributes to a unified perspective on the phenotype, by integrating effects of the environment, genetics, development, physiology, behavior, life history, and evolution.