Please enable JavaScript.

Coggle requires JavaScript to display documents.

PS2021: Mind reading - Coggle Diagram

- - - - Applications
        
        Directly what perceiving
        
        More fine grained information over time & w. features of image\n- Measure of mental workload
        
        Neuromarketing
        
        Mobile eye tracking to see where looking at
        
        Tool for user-centred design
        
        Pupillometry
        
        Measure pupil dilation
        
        Insight into emo state
        
        Limitation
        Not always show emo response
        
        React to diff things
        
        Light intensity/workload/emo
        
        Neuromarketing
        
        Quantitative measures to see what look @
        
        Limitation
        
        Not direct measure of what will buy
        
        User Centred Design
        
        Design easy to use intuitive principles
        
        Useable displays & interfaces
        
        Can use understanding for what drives attention & distraction
      - Mainstream
        
        Used to be lab/marketing setups
        
        Expensive
        
        Gaming laptops often have eye trackers
        
        Webcam based eye tracking
        
        Apple Vision Pro & iPhone
        
        Advantage
        
        Accessibility
        
        Easier user experience
        
        Disadvantage
        
        Give lots additional data to filter through
    - - Types of Tasks
        
        Pro-saccade
        
        Rapidly move eyes toward target
        
        Anti-saccade
        
        Rapidly move eyes in opposite direction to target
        
        Smooth pursuit
        
        Follow moving target w. eyes
      - Measures
        
        Saccade latency
        
        Smooth pursuit speed
      - Significance
        
        Could determine some people based on similar patterns of signature
        
        In principle could match person across diff applications based on eye movement
        
        Participants trained & then matched
      - Limitation
        
        But not clear how generalises to more naturalisitic eye movement
        
        Not clear if would work across wider pop
    - - Features
        
        Saccade length
        
        Fixation duration
        
        No. fixations
      - Significance
        
        Increasing body research using new computer powered techniques can learn individual's eye movement patterns
        
        Recent paper showing eye movement patterns whilst browsing Facebook can tell about personality
        
        Small improvement above change only for some traits
        
        Small diff still hold some useful info when considered across large pop
        
        Could in principle be applied to other traits
      - Limitation
        
        Not clear how generalisable across larger pop
  - - - Locked-In Syndrome (Eickhoff, 2008)
        
        Read mind in situation where can't communicate verbally/make movement
        
        fMRI on coma patient found activity over whole areas
        
        Sig activity in sensory cortices associated w. sensory processing
        
        Activity in language processing areas in response to speech
        
        Familiar voices caused stronger amygdala activation
        
        Linked w. arousal & emo responses
        
        Similar activation as in normal participants
        
        Lack of response attributable to damage in midbrain structures not cortical areas
        
        Advantage
        
        May be able to infer phenomenological experience from physiological response
      - Locked-In Syndrome (Owen, 2006)
        
        2 tasks
        
        1 = Imagine playing game of tennis: x = -2
        
        2 = Imagine walking round all rooms in house starting @ front door: x = -24
        
        When imagine visualised things get brain activity\n- Different activity depending on task
        
        1 = SMA (planning areas)
        
        2 = PMC, PPC, PPA (spatial navigation areas)
        
        Pattern activation of patient & control map onto one another
        
        Limitation
        
        What baseline
        
        But specific thinking does differ
      - Limitation
        
        Reverse inference
        
        Depend how selective region is
        
        Study large brain areas
        
        Unlikely to be only way of characterising
        
        Could have been generated in some unconscious way
        
        No activation doesn't make opposite inference
    - - Supports
        
        Vector machine
        
        Linear discriminant analysis
      - Procedure
        
        Feed lots examples of orientation presented & corresponding brain activity to computer
        
        Computer learns examples
        
        Given new example brain activity computer decodes orientation was likely to be presented
        
        Decoding Seen Orientation (Kamitani & Ting, 2005)
        
        Diff between 2 angles to be decoded
        
        Smaller diff has more overlap of activated neurones\n- All angles measured above chance (50%)
        
        22.5 = 75%
        
        45 = 80%
        
        67.5 & 90 = <95%
        
        Suggest evidence for orientation selectivity in V1
        
        Improvement in diff reflect perceptual diff
        
        Mindreading (Kamitani & Tong, 2005)\n- MVPA decoder trained L/R orientated grating presented separately\n- Discrimination task to detect changes in width of L/R orientated according to central cue\n - 79% correct means v. hard\n- Stimulus changes same only what attending to changes\n - Mind changes\n- Brain patterns caused by attending to L/R grating decoded based on what algorithm learnt from all brain patterns of grating presented on own\n - Mean of all single grating (voxels from areas S1-S4) <95% correct\n - As move up areas % correct decrease where voxels become more specific
        
        Advantage\n- Stimuli in same space means eye tracking not functional\n - Visual input same only participant perception changes\n- Can use activity for actually viewed things\n - To decode purely perceptual images
        
        Limitation\n- Less accuracy than when look @ real image
      - Decode Activity in V1
        
        Respond more to patterns than blank screen
        
        Know from monkey physiology that neurons have preference for certain orientations
        
        Limitation
        
        Neurones equally spaced for orientation preference
        
        No way of getting mean diff
      - Reading Words in Minds (Mitchell et al., 2008)
        
        Meaning of words encoded in terms of how often co-occurs in large text corpus w. set verbs
        
        25 verbs chosen to define meanings
        
        For a set word, verbs learn mapping from intermediate features to MRI images
        
        Predictive model = Stimulus word --> Intermediate semantic features extracted from text --> Mapping learned from fMRI training data --> Predicted activity for verbs --> Combine weightings of co-occurrence data to predict activity for noun
        
        Stats of word occurrence in text can deduce neural representations
        
        Previously only one on mean diff between word categories
        
        Advantage
        
        75% accuracy well above chance
        
        Limitation
        
        Limited set of 60 nouns & 25 verbs chosen by authors
        
        How linked they actually are
  - - - ChatGPT
        
        Use all text on internet as training input to generate text based on input prompt
      - Dall-E
        
        Image generation from text
    - - Critique (Poldrack et al., 2018)
        
        Not mind reading but pattern recognition
- - - - Limitation
        
        No unique physiological response to deception
        
        How to establish baseline
        
        Not account for physiological response to situation in
    - - Readiness Potential (Libet et al., 1983)
        
        Voltage changes in premotor cortex
        
        Activity in ERP signal changes before action made
        
        Can measure intention to act
        
        Readiness potential begins to increase -.05 s before intention spikes & action @ 0s
        
        Significance
        
        See changes in brain before action occur
        
        Even if not carry out action can see intended to
        
        Limitation
        
        Controversial reliability for lie detection & not admissible in court
        
        Hypothesis show neutral then target object and see if change occur when shouldn't
      - P300 ERP (Rosenfeld, 2020)
        
        Event-related potential
        
        Voltage change from baseline @ 300ms after something shown in + direction
        
        Proposed as marker of recog to object
        
        Use for lie detection still controversial