Animal Decision Making - working copy

Currently, this activity of the research group is organised through the research project

"Adaptive Heuretics and Architecture" (AHA!)

financed by FRIMEDBIO in the Research Council of Norway. The AHA! project is also a partner project of Centre for Digital Life Norway. The status of AHA! can be found at our technical website which is updated continuously.

A major activity in the group is to investigate the pathways from (sensory) information to decisions and behaviour in animals. We call this pathway the proximate architecture for decision-making, and we study its effect through evolutionary simulation models.
We first describe this research program, and then our history leading up to it.

Jump to section:

• The proximate architecture for decision-making

• Early work in the research group

• Learning and sociality

• Evolving life histories by genetic algorithms

• The ING method

• Implementing decision modules in simulation models

• Field- and lab-studies

• References

 

The Proximate Architecture for Decision-Making

The model of the Proximate Architecture for Decision-Making (Giske & al, 2013)

We try to develop tools that mimic the proximate architecture for decision-making of an animal (often a fish) and thereby to understand what decisions it makes. The popular science text Oceans of Emotions can give an easy intro to the thinking. A first attempt at this was Giske & al. (2003), but Eliassen et al (2016) and Andersen et al (2016) give a more formal yet strongly simplified model of behavioural architecture and their ecological effects. Giske & al. (2014) shows that this architecture also impacts the genetic diversity of populations, and hence their evolvability.

Core References

 

Early work in the research group

Since 1990 we have been involved in modeling decision-making, particularly in fish and plankton. Much of this work is based on a theoretically derived model for visual range of aquatic organisms (Aksnes & Giske 1993, Aksnes & Utne 1997), which again has allowed calculations of feeding rates and predation risks (Giske & al. 1994, Fiksen & al. 2002).

We have been using Life History Theory (Aksnes & Giske 1990, Giske & Aksnes 1992, Salvanes & al. 1994, Giske & Salvanes 1995, Eiane & al. 1998), Game Theory (Giske & al. 1997) and State-Dependent Optimization (Giske & al. 1992, Rosland & Giske 1994, 1997, Fiksen & Giske 1995, Fiksen 1997, Rosland 1997, Fiksen & Carlotti 1998, Kirby & al. 2000) to model both short-term and life-history decisions, and participated in development and standardization of Individual Based Modelling (IBM; Grimm et al. 2006, 2010, Stillman et al. 2016). More recently, we have been using Genetic Algorithms to evolve adaptive behaviors in IBMs, either directly as life-history decision genes, neural networks of brains, or decisions coming out from the proximate architecture for behavioural control of the individuals. We have also performed field and lab studies to test assumptions in the models.

Core References

 

 

Learning and Sociality

In a series of papers (Eliassen & al. 2006, 2007, 2009) we have studied what conditions favour evolution of learning, and under which circumstances the learning strategy is better or worse than the non-learning alternative. Our key assumption is that individual learning through exploration incurs a time-cost relative to the innate or genetically fixed strategy. Some situations allow coexistence of learners and non-learners in the same population, while life expectancy may be an important determinant for the adaptive value of learning.

Core Reference

 

 

Evolving Life Histories by Genetic Algorithms

The Genetic Algorithm has been a much used research tool in the group for more than a decade. While several papers have used the GA to evolve adaptive values of neural networks or animal decison architectures, the GA can also be used directly to model the major life history decisions in an organism, as done by Fiksen (2000) for the copepod Calanus finmarchicus. This method was also used by Strand & al. (2002) for the mesopelagic fish Maurolicus muelleri.

Core Reference

 

 

The ING Method

The ING is a method for evolving (by a genetic algorithm) flexible adaptive behaviour (controlled by a neural network) in individuals. The background for wanting to develop this tool was the complexity of decisions often faced by organisms. The classical tools in optimization and game are very good at solving specific aspects of adaptive behavior, but by focussing on this single aspect: Life History Theory is a good tool for studies of long-term strategic decisions (and also has been used for the short-term by implicitly assuming constant motivations), Game Theory is good for studies of conflict and cooperation between organisms, and State-Dependent Optimization is good for modelling short-term fluctuations in motivation driven by changes in the (physiological) state of the organism. Through this new method we wanted an agent to be able to consider all these aspects simultaneously (Giske & al. 1998, Huse & Giske 1998).

ING consists of an Individual-Based Model where the decisions in each individual is controlled by its Artificial Neural Network which again is inherited from the parents and evolved by a Genetic Algorithm. Its ability to find the optimal solution has been studied by comparing with dynamic programming (Huse & al. 1999). The tool has been used to model capelin distribution in the Barents Sea (Huse & Giske 1998) and vertical migration in mesopelagic fish (Strand & al. 2002).

However, while ING is able to evolve adaptive solutions to very complex situations, the method does not consider the ability of the organisms to find these solutions. This is the main reason we continued thinking, and arrived at the proximate architecture for decision-making.

Core References

 

 

Implementing Decision Modules in Simulation Models

These theoretical models have also been used in applied models (Fiksen & MacKenzie 2002, Fiksen & al. 2007, Kristiansen & al. 2007, Fouzai et al. 2015), and applied models have been used to compare and test goodness of fit of different behavioural models in applied ecological situations (Vikebø & al. 2007, Kristiansen & al. 2009). We also work on making efficient simplified model formulations (Castellani et al. 2013, Fiksen & Opdal 2015, Sainmont & al. 2015). This is a step from individual behaviour to population dynamics, and maybe a step towards ecosystem and fisheries management tools.

 

Core References

 

 

Field and Lab Studies

In connection with the development of the modelling tools, we have also utilised field and lab studies either to test model assumptions or to study behaviours that models should be able to capture. Our most common field studies include ocean optics, vertical behaviour and life cycles of mesopelagic fishes (Kaartvedt & al. 1988, 2005, Giske & al. 1990, Giske & Aksnes 1992, Balino & Aksnes 1993, Rasmussen & Giske 1994, Goodson & al. 1995, Aksnes & al. 2009, Staby & Aksnes 2011, Staby et al. 2013, Irigoien et al. 2014, Prihartato et al. 2015, Folkvord et al. 2016, Norheim et al. 2016, Røstad et al. 2016) and jellies (Eiane & al. 1999, Ugland et al. 2014, Haraldsson et al. 2014) but also laboratory investigations of vision in fish (Utne & al. 1993, Utne & Aksnes 1994).

 

References