Why do we do the things we do?
It is truly remarkable that much of the awesome variety and complexity of human behavior comes from a somewhat unimpressive grayish organ that weighs in at about 3 pounds!
Our hopes and dreams, our emotions and moods, our personalities, our memories, and our remarkable ability to take actions that we care about are outcomes of electrical and chemical activations in our underwhelming-looking brains.
In our lab, we focus on motivated actions and decision making. Motivated actions are actions we take to pursue outcomes we value. Reaching for a tasty piece of cake is an example of motivated action, and so is avoiding the cake because we have a goal of eating healthy. Stepping away from a snake is an example of motivated action, and so is pursuing a life-time goal of being a herpetologist. Decision making often consists of choosing between two or more motivated actions.
Motivated actions and decisions happen in the brain. But how? How does the brain 'know' how to attach more value to one option compared to another?
The power of the network
One foundational principle of our research is that the amazing capabilities of our brains are the result of millions of distributed computations by neuronal units, often operating in parallel. These neuronal units have a relatively narrow range of functions -- each one cannot be thought of as intelligent in any real sense -- yet working together, they produce complex cognitions and emotions.
A colony of ants can be metaphor for our brains. Each individual ant, just like each individual neuronal unit, is not cognitively sophisticated. Yet, working together, an ant colony can establish highly sophisticated actions -- also via a large number of distributed computations.
Our research seeks to understand motivated action and decision making by developing neural networks that consist of units that interact with each other in ways that are conceptually similar to how neuronal units interact with each other. This approach follows the Parallel-Distributed-Processing (PDP) approach that has heavily influenced psychology and Artificial Intelligence.
The PDP framework is predicated on three simple principles: The first principle is that all cognition takes place in a network that consists of a set of processing elements called units. These units may be connected to each other via weighted connections. Second, processing in such a network occurs via the propagation of activation among the units. The activation of one unit can increase, or decrease the activation of another unit, depending on whether the weight between it and the other unit is positive or negative. Third, the knowledge that governs processing is stored in the weights of the connections, and the effects of experience on information processing are captured by changes to the connection weights.
These principles have been productively applied across an astonishingly broad set of domains in cognitive science. For our lab, it quickly became apparent that the explanatory power of these principles could be productively extended to the domain of decision making. Decisions emerge from patterns of activation in sub-networks. Activation is the sole currency underlying decision making.
The 'readiness' part of a neural network refers to the weights between units, akin to the synaptic strength between neurons. A unit is more likely to fire in response to the activation of another unit if their connection has a large weight. The 'attention' part of a neural network refers to the activation of a representation participating in a decisional context. Readiness and Activation shape Decision-making and motivated action.
A computational model based on a neural network is a tool. It can be used to deepen our understanding of particular decisions/behaviors.
The types of decisions/behavior we are particularly interested include:
1) Contexts in which people don’t act even though those actions may take them closer to goals (Psychological Inertia). For example, many people stay in negative situations, even though taking action could be helpful.
2) Contexts in which people do act despite those actions not being particularly associated with stable goals (Play). For example, people often expend mental effort in doing crossword puzzles because it's fun for them.
3) Contexts that involve affect and affect regulation (e.g. emotion and emotion regulation)
4) Contexts in which people undertake actions that do not appear to be consistent with maximizing value
Recording, analyzing, and modeling such behaviors can shed light on the nature of motivated action