lecture 10 - paradigm shift Flashcards
paradigm shift
- concept by Thomas Kuhn
- a fundamental change in the basic concepts and experimental practices of a scientific discipline
- this happens every so often
paradigm shift: biology
the revolution in thinking about evolution as purposeful to darwin’s idea of natural selection
paradigm shift: physics
newtonian physics was replaced by relativity theory, chaning our idea of space and time
4 stages of a paradigm shift (Kuhn)
- normal science
- paradigm crisis
- adoption new paradigm
- new paradigm dominant
stage 1: normal science
the dominant paradigm is active: a set of theories defines what is possible and how it can be investigated
stage 2: paradigm crisis
- slowly, scientific findings appear that cant be explained well within the current conceptual framework
- stage 2 occurse when the number of difficult-to-explain observations has become too large so that the dominant paradigm enters a stage of crisis
- to combat this crisis, exceptional research is needed to theoretically interpret and experimentally explain the unexplained findings in a different way
stage 3: adoption of new paradigm
a set of new theories and experimental approaches that seem to better solve the identified problems in the future arise
stage 4: new paradigm dominant
- the new paradigm has become dominant
- new textbooks are written
- scientific revolution is complete
paradigm shift: early ideas
- usually not complete nonsense
- old paradigm: limited to a subset of observations
- new paradigm: explains more than the old paradigm (i.e., all observations are explained)
deviant observations
- action signals in primary visual cortex
- 10x as many feedback connections from V2 to V1 than feedforward connection from V1 to V2
- blurry boundaries between cognitive concepts and between cognitive and motor regions
- negligence of the embodied nature of the brain
- theories separate per cognitive function, and no overarching theory
PP as overarching theory
- action, perception, and cognition are united by the same purpose, namely prediction error minimization
how do organisms align the external worlds to internal needs (expectations)
action is particularly important, as it is only through active inference that the brain can align the world to expectations and test model accuracy
what is the first prior
the body
perception as sensorimotor knowledge
- self-movement is necessary for testing and learning sensorimotor dependencies
- sensorimotor knowledge (not sensory knowledge) is a constitutive feature of perception
WM properties
- it is for the future
- it is action-oriented
Phylogenetic refinement to define brain functions
- approach is grounded in phylogeny, the evolutionary history of species
- implies that brain functions should be understood through the lens of control of interactions with the environment (e.g., grasping or reaching for an object)
- focuses on pragmatic representations.
limits to the cognitivist approach
- lack of transfer of learning in computerized cognitive training
- deep trouble in AI, which takes a similar data-driven learning approach
- i.e., stimulus driven learning does not lead to transfer of learning
How to induce general learning at the cognitive level
- top-down learning: this is necessary for learning at higher-more abstract levels. this also involves active inference.
- learning in more variable and naturalistic settings
generic vs context-bound cognitive skills
- generic skills: are applicable across contexts
- context-bound skills: are tied to specific situations
- situated cognition and enactivism emphasize that learning is embedded in action and context. This suggests that some cognitive skills may only develop in specific environments where action and perception are coupled
conceptual shifts
- predictive processing
- enactivism/action-oriented cognition
- phylogenetic refinement
methodological shift
studies in more naturalistic settings in which subjects (or deep neural nets) can self-act on more naturalistic stimuli
reductionism in neuroscience
- modern cognitive neuroscience relies heavily on reductionism
- uses simplified stimuli and tasks that neglect the importance of actions as a critical factor in survival and cognition
- 2D images cannot be acted upon and dont serve needs
theoretical reasons for moving away from proxies
- actability: cognition is ultimately geared toward action and survival
- realism vs. symbolism: 2D images are representational and often symbolic, so they lack the importance of real world presence. they dont satisfy needs like real objects
- multisensory processing: in the real world, we don’t just rely on one sense (e.g., vision) to interact with objects; we use multiple senses
- motion and depth: 2D images lack the ability to convey motion and depth, both of which are essential for understanding the real world size and actability of objects
- all these factors are important for normal development
evidence for differences
- real-object advantage: better recognition performance and memory for real objects, due to actability
-
real-object preference:
- capture gaze and attention more
- show reduced repetition suppression, meaning our brain reacts more robustly to repeated presentations of real objects
- certain neurons only fire to objects within reach since they allow interaction (actability)
- real food is more valuable than an image of food
qualitative differences
- real vs artificial stimuli evoke different patterns of behavior and neural activation
- this suggests that previous conclusions about behavior and brain processing may not generalize to real stimuli
qualitative differences
- 2D images provide far less information about physical properties like size and weight
- because they lack the actability and tangible qualities that allow for a richer, more direct interaction with real or 3D objects.
build-up and tear-down approaches
- to test how findings from controlled, artificial lab settings can be generalized to more complex, naturalistic environments
- build-up approach: begins with minimalist stimuli and progressively adds complexity to understand which components contribute to realistic cognition (reduced → real stimuli)
- tear-down approach: starts with real-life, complex settings and removes aspects to observe which components are most critical for realistic cognitive processing (real → reduced stimuli)
when can we develop valid, real-world applications
when there is no difference in the effect of realistic and proxy stimuli on a technological or theoretical level
if the source of difference between reality and proxy is technological
- work with industry to improve tech accordingly
- compare reality to proxy
- if there is no difference, develop valid real-world applications
if the source of difference between reality and proxy is theoretical
- understand the factors yielding the difference
- develop real world applications
methodological approaches in neuroscience
REDUCTIONIST
- reductionism: essential approach using simplified/reductionist methods to help break down complex phenomena into more manageable parts for study
IMMERSIVE NEUROSCIENCE APPROACHES
- build-up approach
- tear-down approach
- studying behavior and cognition in their full complexity within natural environments, instead of experimental dissection of behavior and brain processing
immersive neuroscience
- an approach that seeks to study cognition and brain function in more naturalistic, real-world settings, as opposed to relying solely on simplified, reductionist laboratory conditions
- does not mean the reductionist approach is invalid
challenges in immersive neuroscience
motion artifacts: The major techniques used to record brain activity—M/EEG and fMRI—are highly sensitive to any movement during data collection
methodological advances
- better methods for removing M/EEG artifacts
- scalp EEG with backpack equipment
- invasive EEG - New generation of MEG sensors
- Optically Pumped Magnetometers (OPM)
optically pumped magnometers (OPM)
new generation of MEG sensors that can be placed directly on the head
OPM advantages
- more mobile: participants can perform tasks while their brain activity is being recorded
- much less sensitive to motion artifacts: effective in capturing brain activity even when participants are moving, which is a significant step forward for studying cognition in dynamic, real-world environments
- much cheaper than regular MEG
- high spatial and temporal resolution
OPM disadvantages
- sensitive to earth’s electromagnetic field, so it needs a shielding chamber
- doesnt record subcortical activity
less-traditional experimental approaches
- iEEG
- scalp EEG + backpack
- OPM
- wearable technology with greater reliance on AI
ecological validity vs experimental control trade-off
- As ecological validity increases, experimental control tends to decrease. Real-world stimuli are harder to manipulate and control precisely, making it difficult to isolate specific variables.
- On the other hand, as control increases, the conditions become more artificial and may not generalize well to real-world scenarios.
