bubic2010: Prediction, cognition and the brain

tags
Predictive Processing, Brain
source
https://www.frontiersin.org/articles/10.3389/fnhum.2010.00025/full

This paper acts as an introduction to the vast literature considering the brain as a predictive processing machine. Specifically, it is looking at terms which have been introduced such as prediction, prospection, anticipation, expectation, preparation, as well as violations of expectations or prediction errors. All these terms have different nuances and levels of abstraction associated with human behavior, and should be treated separately.

The Predictive Brain

Ideomotor Principle

We start our story with the ideomotor principle, which purports a shared (or common) code (neurological code that is) between perception and action. What this is saying is that action and perception (our senses) are intrinsically linked in the core of human processing. Several papers and reports have revisited this idea such as

  • (Prinz 1990) Proposes a common coding for action and perception
  • (Stock 2004) Which reviews the ideo-motor action principle in both a modern and historical context.

It is important to note that in the discovery/formulation of the ideomotor principle, James (one of the major proponents) casted a key component, that of sensory anticipation, as a “pre-perception” – this idea goes to lay the ground work as sensory predictions and anticipation in future predictive processing work.

Helmholtz

Another early influence in predictive processing, and really representationalism as we know it today, is von Helmholtz who laid the ground work for learning representations of sensory signals through “sign coding”. The major breakthrough here was the idea that processing in the brain was not an a priori process, but a learned process. Specifically, the brain makes small changes and adjustments to sensory input based on minute inferences and the abstract inference of the overall system. This idea would be further developed into ideas pertaining to error correction and backpropagation in artificial neural networks.

Concern with top-down-processes

  • Predictive approaches tend to focus on top-down processes, focusing on how prior knowledge influence future behavior.

  • Many processes are studied in isolation, where researchers consider independent parts of the architecture w/ some process feeding in features through the formation of new features which are then fed to higher order cognitive functions. While this may be good for uncovering information about specific processes, this isolation may lead researchers in the wrong direction.

  • One important concept that was introduced here is the distinction of memory and prediction, which had thought to be synonymous prior. (Soga 2009) shows that the brain may employ predictive and memory capacities in different contexts.

Many Terms, How Many Meanings?

  • Anticipation or Preparation (LaBerge 1995): elevated levels of processing in sensory or motor areas occurring prior and facilitating the processing of the expected perceptual or motor event.

  • Expectation (LaBerge 1995): reflects a memory component as it refers to an item stored in either working or long-term memory which includes the information regarding the spatial and temporal characteristics of the expected event.

    Because the representations discussed in the definitions above can be coded in rather abstract/verbal forms they do not necessarily presuppose a pre-activation of the relevant sensory cortices.

  • Prediction (Butz 2003): refers to a representation of an event (potentially comparable to the LeBerge’s definition of expectation).

  • Anticipation (Butz 2003): describes the impact of predictions on current behavior (decisions and actions based on “anticipatory” signals).

  • Prospection (Gilbert 2007): ability to “pre-experience” the future by simulating it in our minds, which may lack the detail and richness of genuine perceptions.

This term is loosely defined and contains some aspects of expectation and anticipation. It is unspecified in which extent and under what conditions anticipation/expectation are evoked. Prospection may be better suited for the general orientation towards the future in a sense that stored information is constantly used to imagine, simulate, and predict future events.

  • Prospective codes ({Sch{\“u}tz-Bosbach} 2007): event production and simulation as representations of present events which contain information pertaining to their future effects or goals.

    The previous definitions have traditionally regarded predictions as forms of attention. It recently has been suggested that expectations/predictions represent fully distinct phenomena (Summerfield 2009). When comparing the terms prediction and attention, it may be of use to clearly specify the aspect of attentive processing to which predictive processing is being compared.

    Bubic suggests that “predictive processing” should be used for describing the general orientation towards the future which includes a wide range of predictive phenomena.

Nature and strength of prediction

The nature and strength of predictions varies greatly in different contexts and may be influenced by different factors.

One such separation between implicit anticipations expressed through habits (behavior) and explicit ones which include representations of the predictive future states is expressed in (Pezzulo 2008). It is unclear if supporting a dichotomy between these two types of predictions is worthwhile, and really I expect there to be a continuous distribution (as suggested by Bubic).

Different Temporal Scales:

  1. expectations can be formulated based on the knowledge gained through long-term experience (Bar 2007), or learning triggered by short-term exposure to non-random patterns (Schubotz 2007).
  2. It is possible to predict events which are expected to occur in different moments in the future (e.g. those expected to occur within seconds-range in contrast to those which may occur in the distant future).
  • Long-term predictions can be thought to be used offline, and not coupled with any immediately relevant process. (less accurate)

  • Short-term predictions can be thought to be used online, and directly tied to the regulation of ongoing behavior. (more accurate)

Prerequisites and benefits of prediction

Prerequisites of prediction

Events can be predictable if they occur in a non-random fashion, allowing the brain to extract either deterministic or probabilistic regularity of the relationship between different events.

What makes an event more predictable than another? Are there ways we can predict an event which are “more learnable”?

  • First-order rules defined by the repetition of a stimulus (stimulus feature) can trigger an expectation about the continuation of its appearance in the future (Squires 1976)
  • Second-order (or even higher order) (contingency) rules which require the extraction of relations between specific, mutually non-interchangeable stimuli can underlie expectations related to more complex events, even those which were previously not encountered within the respective context.

Benefits of prediction

  • Wundt showed that attention and expectations related to upcoming stimulus can shorten perception time (LaBerge 1995)
  • Lange showed beneficial behavioral effects following the correct anticipation of a response (LaBerge 1995)
  • Expectations allow us to construct a coherent and stable representation of the environment which is usually not easy, given the available, often impoverished information (Kveraga 2007)
  • Predictions may guide top-down deployment of attention, improve information seeking as well as subsequent decision making (Butz 2008)

More generally, the ideomotor principle suggested that anticipations of consequences to actions can trigger and guide behavior. This effect has been shown a number of times, specifically by studies showing that representations of events or actions include the anticipated effects of those events. (Kunde 2007) argued that anticipation is necessary for action. This is because any response must be initiated by a response-related anticipation.

Mechanisms of Prediction

Anticipatory/predictive processing is directed towards the future, yet is highly dependent/grounded in past experience. While this is the case, it is hard to identify common neural mechanisms supporting such processing across all contexts.

Sources and sites of predictions in perception

Conceptualize anticipation as a bias signal (Rees 1998) to improve computational efficiency of a specific area. This may be useful as a means to specify three elements which need to be specified:

  • brain regions which formulate expectations and impose such a bias (sources),
  • regions which are influenced by predictions (sites),
  • and a communication mode mediating this process.

While conceptually we can draw clear lines between these neuronal groups, the separation may not be as clear in the brain. For example, the predictive coding model gives an account of a hierarchical approach where top-down “sources” influence bottom-up “sites” which in turn are sources for lower level sites. This means that the higher level predictions influence lower level predictions.

Internal models mediate prediction across many domains

It has been suggested that the prediction of future states of the body or the environment arises from mimicking their respective dynamics through the use of internal models.

What is an internal model? A “dynamical model” which simulates the dynamics of the system/embodiment.

The brain as a prediction device

If one was to try to summarize all brain areas which have been mentioned as incorporating some aspect of predictive processing, these would include the whole (or at least most of) brain.

References

(Prinz 1990) W. Prinz, A {{Common Coding Approach}} to {{Perception}} and {{Action}}, (1990).

(Stock 2004) Armin Stock and Claudia Stock, A Short History of Ideo-Motor Action, Psychological Research, , pp. (2004). .

(Soga 2009) Ryosuke Soga; Rei Akaishi and Katsuyuki Sakai, Predictive and Postdictive Mechanisms Jointly Contribute to Visual Awareness, Consciousness and Cognition, , pp. (2009). .

(LaBerge 1995) David LaBerge, Attentional Processing: {{The}} Brain’s Art of Mindfulness, (1995).

(Butz 2003) Martin V. Butz; Olivier Sigaud and Pierre G{\‘e}rard, Anticipatory {{Behavior}}: {{Exploiting Knowledge About}} the {{Future}} to {{Improve Current Behavior}}, (2003).

(Gilbert 2007) Daniel T. Gilbert and Timothy D. Wilson, Prospection: {{Experiencing}} the {{Future}}, Science, , pp. (2007). .

({Sch{\“u}tz-Bosbach} 2007) Simone {Sch{\“u}tz-Bosbach} and Wolfgang Prinz, Prospective Coding in Event Representation, Cognitive Processing, , pp. (2007). .

(Summerfield 2009) Christopher Summerfield and Tobias Egner, Expectation (and Attention) in Visual Cognition, Trends in Cognitive Sciences, , pp. (2009). .

(Pezzulo 2008) Giovanni Pezzulo, Coordinating with the {{Future}}: {{The Anticipatory Nature}} of {{Representation}}, Minds and Machines, , pp. (2008). .

(Bar 2007) Moshe Bar, The Proactive Brain: Using Analogies and Associations to Generate Predictions, Trends in Cognitive Sciences, , pp. (2007). .

(Schubotz 2007) Ricarda I. Schubotz, Prediction of External Events with Our Motor System: Towards a New Framework, Trends in Cognitive Sciences, , pp. (2007). .

(Squires 1976) K. C. Squires; C. Wickens; N. K. Squires and E. Donchin, The Effect of Stimulus Sequence on the Waveform of the Cortical Event-Related Potential, Science, , pp. (1976). .

(Kveraga 2007) Kestutis Kveraga; Jasmine Boshyan and Moshe Bar, Magnocellular {{Projections}} as the {{Trigger}} of {{Top}}-{{Down Facilitation}} in {{Recognition}}, Journal of Neuroscience, , pp. (2007). .

(Butz 2008) Martin V. Butz and Giovanni Pezzulo, Benefits of {{Anticipations}} in {{Cognitive Agents}}, (2008).

(Kunde 2007) Wilfried Kunde; Katrin Elsner and Andrea Kiesel, No Anticipation\textendash No Action: The Role of Anticipation in Action and Perception, Cognitive Processing, , pp. (2007). .

(Rees 1998) Geraint Rees and Christopher D Frith, How Do We Select Perceptions and Actions? {{Human}} Brain Imaging Studies, Philosophical Transactions of the Royal Society of London. Series B: Biological Sciences, , pp. (1998). .