Cognitive mapping: Introduction and its role in creating neural networks assisting in group training

Introduction to cognitive mapping and the associated neural networks

Since the 70s, ongoing hypotheses, theories, and evidence are supporting the knowledge and application of the Cognitive map theory. The initial contribution towards this theory came from O’Keefe & Nadel. They hypothesized that the hippocampus contains a spatial map of the environment, as a result of research done on rats in enclosed environments. This research concluded that certain neurons responded really strongly when the rat was in a certain position. These neurons are called Place cells and respond when an organism is in a specific location in an allocentric space – A representation of a spatial environment referenced to an external coordinate system that is not dependent on the view or direction navigated.

This latter mentioned, led to the theory that a collection of neurons could function as a map of the particular environment an organism is in. However additional research has shown that these place cells do not isolate the data of the location, the responses are highly context-dependent. Due to these findings, place cells are seen as entities that integrate space within contextual cues – something that is of great importance for memory in general.

In the Morris water maze study, conducted in 1982, the hippocampal storage of spatial mapping is tested by including rats with lesions to compare their abilities with control rats. The rat with hippocampal lesions needed to rely on trial-and-error while the control group was able to accomplish their tasks due to their cognitive mapping. This is shown in figure 1.

Fig. 1. The route taken by a typical rat in the Morris water maze. The control rat and ones with cortical lesions can remember the location of the submerged platform and go directly there, whereas the hippocampal-lesioned rats find the platform by trial and error. Adapted from Morris et al., 1982.

Moving on from rat studies, cognitive mapping is also seen in humans. Various functional imaging and lesion studies help to improve the knowledge regarding cognitive mapping in people. In contrast to rats, humans do possess higher amounts of lateralization in hippocampal functioning. Especially, the right hippocampus is needed to obtain the information to achieve spatial memory while the left hippocampus is needed to retrieve the contextual information.

In addition to the place cells, there are more regions in the medial temporal lobes that contribute to spatial awareness. Instead of single location responses, there are cells that respond to multiple locations in a repeating, triangular, and grid-like manner. These cells are called the Grid cells and help to form repeating grid-like patterns that complement the contextual spatial information of the Place cells. These Grid cells may possess the ability to assist in linking visuospatial signals to other cognitive abilities to perform better in a given spatiotemporal context.

Evaluating a model for hierarchical behavior through cognitive mapping

The model of Jordan et al. proposes a neural architecture that integrates cognitive mapping into goal-orientated actions. This model is made through the cognitive processing lens of an agent, or an individual acting in a certain situation. The agent gets constant input from the state cells that process information coming from the sensory input (Look at our article Cognitive Load Theory to learn more about information processing). Additionally, there is a constant flow of information coming from the goal cells.

Given these information inputs, the model is expanded by including the state-action and sequence cells that work together towards actions that will enable the agent to work towards their goals. Besides, the gate cells make sure to process the planned action and make sure that the relevant plans will pass through to the action cells. These latter mentioned, help the agent to act. This is all seen in figure 2.

Figure 2. The architecture of the hierarchical neural network model. Adopted from Jordan et al., 2020.

Zooming in on the state-action (SA) cells

The research of Jordan et al. goes on to look at the state-action cells. Since these cells are capable of responding to the combination of both state and action information the theory is that these cells should be located in an area with high levels of motor as well as sensory feedback. One hypothesized location can be found in the prefrontal cortex – which is located in the right part of the sensorimotor pathway. Additionally, all kinds of sensory modalities activate the prefrontal cells upon stimulation.

This is valuable for the cognitive abilities of individuals since these cells contribute to the anticipation of certain events as well as help responses before and during activities to act upon the anticipated behavioral consequences. Besides, High-level actions do need prefrontal neurons to fire up the premotor cortex. As discussed in our Executive functioning article, the prefrontal cortex is of fundamental importance to perform executive function at the top of the motor hierarchy. Lesion studies supported these theories. In performing a task, predominantly primate prefrontal cortex cells respond to sensory cues and act upon motor actions through associative learning.

In addition, there is recurrent connectivity identified between state-action cells, anticipation, and behavioral consequences. This can be visualized in a backward causal and a forward causal model, representing states-action combinations based on previously achieved results and predicted state-action combinations to plan future actions respectively. The relevancy of both models is high since these determine what kind of information is processed and what kind of information is forgotten (see Cognitive Load theory article).  These models are shown in figure 3.

Fig. 3. Connectivity patterns between SA cells encoding for a backward causal model. Adapted from Jordan et al., 2020.

Cognitive Mapping in group training

Curseu et al. obtained results from conceptual mapping techniques in relation to developing effective group training strategies. In various studies, the trainability of group cognition has been stated as a way of improving group performance. In order to achieve cognitive changes, cognitive mapping is a topic that needs exploration in order to induce structural change. In doing so, and in providing the background for the identification, evaluation, and training of group cognition, tools can be offered to improve the performance.

Additionally, Curseu et al. are in support of using cognitive mapping to help members reach a certain consensus as well as assist in diagnosing and managing disagreements that come up under great stress, pressure, and uncertainty. In certain training settings, this ability helps create tools to help group members acknowledge their problems in order to find effective ways to perform at the highest level regardless of the stress, pressure, and uncertainty they need to deal with. 

References

  • Cole, M. A., & Ward, J. (2010). A Student’s Guide to Cognitive Neuroscience. Journal of the International Neuropsychological Society: JINS16(5), 945.
  • Curşeu, P. L., Schalk, R., & Schruijer, S. (2010). The use of cognitive mapping in eliciting and evaluating group cognitions. Journal of Applied Social Psychology40(5), 1258-1291.
  • Hafting T, Fyhn M, Molden S, Moser M-B, Moser EI. 2005 Aug. Microstructure of a spatial map in the entorhinal cortex. Nature. 4360(7052):801–806. 
  • Hasselmo ME. 2005. A model of prefrontal cortical mechanisms for goal-directed behavior. J Cogn Neurosci. 170(7):1115–1129. doi:10.1162/0898929054475190. 
  • OC Jordan, H., Navarro, D. M., & Stringer, S. M. (2020). The formation and use of hierarchical cognitive maps in the brain: A neural network model. Network: Computation in Neural Systems31(1-4), 37-141.
  • Morris, R. G., Garrud, P., Rawlins, J. A., & O’Keefe, J. (1982). Place navigation impaired in rats with hippocampal lesions. Nature297(5868), 681-683.