Thinking before doing ?
Mental representations
“Any mental content or operation that stands for something else in the world”. → categories,
symbols, mental images, memories etc.
Function:
- They help us understanding the world around us. We use our mental representations and
then observe the world around us.
- Classification → classifying humans and objects
- Additional attributes → what can we do with these objects ?
- Steering attention → when you are thirsty you will be able to detect glasses of water faster
- Communication/thinking → mental representations help communicating/thinking about
objects
Availability & accessibility:
- All kind of knowledge is available to us in our mind. There is only a limited part of the
available information that is accessible to us.
- “Accessibility can be defined as the activation potential of available knowledge.”
Activation of mental representations (research) → authors hypothesized that after watching the
Chinese woman putting on make-up, the social category of females is more accessible than the social
category of Chinese people. And the other way around for the group seeing the woman eating with
chop sticks.
- Faster response latencies are referred to more accessibility of that specific information. After
watching the make-up video, people are faster in categorizing words that are related to
women. They were slower doing that in the chop-stick condition.
Storing knowledge
,Associative network models:
- (A-modal model) mind works like a computer. As soon as I see coffee, there are all kinds of
notes that can become active by spreading activation (awake, black etc.).
- Links are formed through experience. The strength of links is thought to be determined by
the frequency and recency of activation.
- It also provides an explanation for priming → activating a concept with a prime leaves it
momentarily higher in activation and ready to be engaged in subsequent processing.
- Spreading activation is typically treated as unconscious. People become aware of the
concept that results from nodes that have reached the threshold of consciousness.
Parallel search → Activation spreads automatically among all connected nodes (unconsciously until
reaching threshold and then the concept comes to mind and can be used explicitly)
Serial search → the nodes and pathways are themselves unconscious, except for the node that is the
present focus of the serial search (the search process itself is thought to be conscious)
- Dual-process models of representation suggest that people may use both structures, with
parallel search as implicit and serial search as an explicit search model
Schema models:
- Schemas provide understanding of those connections between the features of a specific
object and provide rules of the situation (how to make a cup of coffee). Once a schema is
activated, it operates as a lens, directing your attention, memory and judgement.
- Unlike the bottom-up dynamics of associative network models, schema models take a top-
down approach: large chunks of general knowledge might be thought of as different lenses
with which to view the world. Schemas are activated and used one at a time. Once activated
a schema brings online all of its general knowledge → directs attention, memory and
judgment
- Although the content of the traits, stereotypes and other schemas may be available to
consciousness, their activation and application in thinking and judging are presumably
unconscious.
Predictive coding:
- Bayesian processes: prior knowledge is used in perceiving the world around us. My prior of a
tasty cup of coffee was violated as the coffee was bad, so you would update your prior
related to coffee at this specific bar to the posterior (comparison between perception and
prior). Schemas can be predicted in predictive coding as a “prior”.
Connectionist models:
- Mimic the network structure of human neurons
- (A-modal model) besides the facilitative links (one note activating another note), there are
also links that inhibit activation of other nodes. The full set of nodes that is activated make a
mental representation and therefore these representations are not static but affected by the
environment.
- Like associative network models, there are multiple processing nodes at varying levels of
activation.
- Most connectionist models assume that a specific node doesn’t have semantic meaning
- There is no single distributed representation for a concept → each time the nodes and
weighted connections will reach a slightly different approximation based on the combination
of prior experiences and other contextual influences
, - Localist connectionist models → single nodes can have semantic meaning (one neuron can
represent specific things, places or persons). Some argue that this is more similar to
associative network models than connectionist models.
- Distributed representations might be described as efficient and unintentional and people
become aware of the final emergent pattern, but not the workings of the connectionist
network itself. Current research seems to assume that connectionist models are primarily
models of implicit representation.
Connectionist models account for commonly observed effects:
- Recency effects → recently instantiated patterns make the connectionist model more likely
to fall into similar patterns for a certain period
- Chronic accessibility → slowly learned changes in weights between nodes make certain,
strongly connected concepts arise easily
- Schematic processing → triggering one portion of a given pattern makes it easier for the rest
of the pattern to complete
Multiple format models:
- Amodio criticizes the idea that representations are created and stored in the same way.
Within the memory system model there are multiple learning systems that are related to
different memory systems in different brain regions (affective, semantic and procedural
memory system). The memory system model argues that learning and behavior are linked
based on neurological pathways.
Embodied cognition:
- Representations are modality-specific → (Modal model) representations are constituted of
sensory experiences (you drink coffee and experience sensory experiences and all these are
included in the mental representation). Just by thinking the representation to some extent,
you activate these sensory experiences (partial re-experience so actually have the feeling of
smelling coffee)
Amodal forms of mental representation → abstract and disembodied symbols of objects or events
that do not retain the sensory components from the original experiences of those objects or events
(think about coffee as warm, arousing, but in an abstract and conceptual way, that does not actually
call upon sensory information or details).
Modal forms of mental representation → are constituted by those sensory experiences: heat,
flavor, alertness.
- For amodal models, the modality-specific information is translated into an abstract,
disembodied format for further use.
- In embodied (modal) models, details of body matter for details of mind. Embodied
information is not translated into abstract currency, but itself is the basis of mental
representation. Mirror neurons are an example for this.
Situated cognition models:
- Mental representations result from dynamic interactions between the brain, body, and
environment. So while the first models mainly thought of mental representation computer-
like, later models included the body or even the situation when thinking about mental
representations.
, - The brain can delegate to features of the environment (phone etc.) and simplify decision
making.
Models differ on whether the representations are themselves conscious.
Models agree on that the concept, idea or thought that results from activated representations
could become conscious under the right circumstances.
Automaticity
Four horsemen of automaticity:
- Level of consciousness → if things happen outside conscious awareness and it still has an
effect
- Level of intentionality → when we do things that are not in line with our goals/intentions, it
is considered more automatic (mistakenly going into wrong room, because you’ve worked
there before. So, the behavior was non-intentional).
- Level of controllability → “you cannot kill yourself by holding your breath”. That shows the
strength of an automatic process.
- Level of efficiency → by time you learn how to efficiently do things (doing a lot of things at
the same time)
- All those four features are independent features, which can be high or low
Impulsive vs. reflective behavior
Humans do not always behave in line with their values. Several strategies have been pursued:
1. Assume ignorance or lack of knowledge of the actor
2. Behavior may occur mindlessly or automatically
3. Basic drives may override considerations of utility and immediately determine behavior
Strack & Deutsch reflective impulsive model:
Thesis 1:
- Basic assumption → social behavior is the effect of the operation of two distinct systems of
information processing (reflective and impulsive)
- Impulsive processes → Behavior within the impulsive system is based on spreading
activation of knowledge to motor representations according to ideo-motor processes.
Thinking about behavior is enough to trigger the behavior itself (William James 1890).
- Reflective system → there is a weighting of the consequences of several behaviors
Thesis 2:
- Parallel operation → both systems operate in parallel. Impulsive system is always activated
and the reflective system is sometimes activated whenever there is enough motivation and
opportunity.
Thesis 3:
- Capacity → reflective system requires a high amount of cognitive capacity and distraction or
high levels of arousal will interfere with its operation. Impulsive system requires little
cognitive capacity and may control behavior under suboptimal circumstances.
Thesis 4:
- Relations between elements → In reflective system, elements are connected through
semantic relations to which a truth value is assigned. In the impulsive system, the relations
are associative links between elements and are formed according to principles of contiguity
and similarity.
- The accessibility of content in the impulsive system will be increased by the frequency and
recency of prior activation → consequence are associative clusters of frequently occurring
features.
