1 / 30

Being Virtual: Consciousness and self as graded, adaptive phenomena

axel cleeremans. Being Virtual: Consciousness and self as graded, adaptive phenomena. Cognitive Science Research Unit http://srsc.ulb.ac.be/. SRSC. The issues. In what sense can cognition be unconscious?

eustacia
Download Presentation

Being Virtual: Consciousness and self as graded, adaptive phenomena

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. axel cleeremans Being Virtual:Consciousness and self as graded, adaptive phenomena Cognitive Science Research Unithttp://srsc.ulb.ac.be/ SRSC

  2. The issues • In what sense can cognition be unconscious? • Despite increasingly widespread recognition of the phenomena of implicit cognition, the issue continues to be hotly debated in cognitive psychology: • Important methodological problems need to be solved • Brain imaging techniques do not solve all these problems • The core of the debate concerns the possibility of unconscious representation • What is the function of consciousness? • Very few existing computational proposals • Most contemporary theories of cognition have ignored the issue Banbury — 2

  3. A B C D E F The sequence learning paradigm • Choice reaction • Stimulus movements follow a … simple repeating sequence • Learning is assessed by switching to a different sequence during a transfer block DA C B E F D A C B E F … (training) F B E A C B F B E A C B … (transfer) Banbury — 3

  4. Method • training: • 15 blocks of a 4 choice reaction time task using a repeating 12 elements SOC sequence: a b c d a c b a d b d c • 13th block on another SOC sequence. • 2 conditions: 0 msec RSI vs. 250 msec RSI • Generation task: • after the RT task, participants are asked to generate a 96 trials sequence under inclusion and exclusion instructions, without feedback. • inclusion: try to reproduce the training sequence • exclusion: try to generate a sequence that is different from the training sequence • Recognition task: • subjects were presented with 24 short three trials sequences and had to classify them as familiar or novel by rating each on a 6 points scale. Banbury — 4

  5. Serial Reaction Time Task Results • Participants trained with a 250 msec RSI learn faster and respond faster than participants trained with a 0 ms RSI • Participants learn about the sequential structure in both conditions • Participants trained with a 250 ms RSI exhibit a larger (but non-significantly different) transfer effect than 0 ms RSI participants Mean RT (ms) 500 480 460 440 420 400 380 360 340 RSI 320 no RSI 300 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 Training blocks Banbury — 5

  6. Generation no RSI scores RSI 0.6 0.5 0.4 0.3 0.2 Inclusion Generation Task Results • For each participant, we computed the number of generated chunks of length 3 (max = 94) that were part of the training sequence, in both inclusion and exclusion conditions • Inclusion scores exceed baseline (33%) in both conditions • Most studies stop here and would conclude that learning was in fact explicit Banbury — 6

  7. Generation no RSI scores RSI 0.6 0.5 0.4 0.3 0.2 Inclusion Exclusion Generation Task Results • RSI participants can do the difficult exclusion task! • Exclusion scores exceed baseline in the no RSI condition only • Subjects have no control on their knowledge of the sequence in the no RSI condition • Subjects have learned the sequence in the no RSI condition but their knowledge appears to be essentially implicit Banbury — 7

  8. Recognition scores old new 4.00 3.50 3.00 2.50 2.00 no RSI RSI Recognition Task Results • Participants in both conditions do not respond faster to old than to new sequences. • No effect of perceptual fluency in the recognition task. • Subjects are able to differentiate between old and new sequences of three elements in the RSI condition but not in the no RSI condition (p < 0.05) • Participants in the no RSI condition lack explicit knowledge about the sequential regularities Banbury — 8

  9. The plan • Being a zombie: A brief overview • Being Conscious: A framework • Being Virtual: A speculation Banbury — 9

  10. > Being a Zombie • Considerable evidence for unconscious cognition and learning • Learning by foetuses, memory consolidation during sleep, role of experience in shaping basic neural maps (penfield homonculus) • Implicit learning, subliminal priming • Learning in amnesia, blindsight Banbury — 10

  11. Commander Data meets the zombies • Two strategies to account for the existence of a cognitive unconscious: • Zombie theories assume total duplication of functions: The cognitive unconscious is just like the cognitive conscious, only minus consciousness • Commander Datatheories assume that mental life is co-extensive with consciousness: Whenever some state is representational, it is also a conscious state • Both accounts are rooted in the “classical” notion that cognition consists of symbol manipulation • Both accounts fail to offer consciousness a clear computational function. It is a pure epiphenomenon Banbury — 11

  12. >> Being Conscious • Explore an alternative framework in which: • Conscious and unconscious processing are rooted in the same basic mechanisms • Consciousness is a graded, continuous, and dynamic phenomenon • Assumptions about: • Processing (P1-P4) • Representation (R1-R3) • Learning (L1-L3) • Consciousness (C1-C5) • Self (S1-S7) Banbury — 12

  13. Assumptions about processing P1: The cognitive system is best viewed as involving a large set of interconnected processing modules organized in a loose hierarchy. Each module in turn consists of a large number of simple processing units connected together P2: Long-term knowledge in such systems is embodied in the pattern of connectivity between the processing units of each module and between the modules themselves P3: Dynamic, transient patterns of activation over the units of each module capture the results of information processing conducted so far P4: Processing is graded and continuous: Connected modules continuously influence each other’s processing in a graded manner that depends on the strength of the connection between them and on the strength of the activation patterns that they contain Banbury — 13

  14. Assumptions about Representation R0: [Representations are necessary as mediating states] R1: Representations consist exclusively of the transient patterns of activation that occur in distributed memory systems R2: Representations are graded: They vary on several dimensions that include strength, stability in time, and distinctiveness R3: Representations are dynamic, active, and constantly causally efficacious Banbury — 14

  15. Assumptions about Learning L1: Adaptation is a mandatory consequence of information processing • LTP & LTD, hebbian learning L2: Learning is adaptation that specifically involves high-quality representations • Distinction between weight-based learning and activation-based learning L3: Learning has both direct and indirect effects Banbury — 15

  16. Assumptions about Consciousness C1: Consciousness involves two dimensions: Subjective experience and control C2: Availability to consciousness correlates with quality of representation C3: Developing high-quality representations takes time C4: The function of consciousness is to offer flexible, adaptive control over behavior C5: Learning shapes conscious experience Banbury — 16

  17. implicit cognition priming explicit cognition automaticity Potency Availability to A-Consciousness QUALITY OF REPRESENTATION (stability, distinctiveness, strength) Processing is graded R2:Representations are graded: They vary on several dimensions that include strength, stability in time, and distinctiveness Banbury — 17

  18. implicit cognition priming explicit cognition automaticity Potency Availability to A-Consciousness QUALITY OF REPRESENTATION (stability, distinctiveness, strength) Processing is graded R2:Representations are graded: They vary on several dimensions that include strength, stability in time, and distinctiveness R3: Representations constantly influence information processing regardless of their quality Banbury — 18

  19. implicit cognition priming explicit cognition automaticity Potency Availability to A-Consciousness QUALITY OF REPRESENTATION (stability, distinctiveness, strength) Processing is graded R2:Representations are graded: They vary on several dimensions that include strength, stability in time, and distinctiveness R3: Representations constantly influence information processing regardless of their quality C3: Developing high-quality representations takes time, both over learning and development, and within a single trial Banbury — 19

  20. implicit cognition priming explicit cognition automaticity Potency Availability to A-Consciousness Availability to Control QUALITY OF REPRESENTATION (stability, distinctiveness, strength) Control and the function of consciousness • Weak (“implicit” ) representations do not need control because they only exert weak effects on behavior (I.e., through priming) • Strong (“automatic”) representations do not need control because they are adapted (L1:Adaptation is a mandatory consequence of information processing) C4:The function of C is to offer flexible, adaptive control over behavior (over those representations that need most control because they drive behavior) Banbury — 20

  21. implicit cognition priming explicit cognition automaticity Availability to P-Consciousness Potency Availability to A-Consciousness Availability to Control QUALITY OF REPRESENTATION (stability, distinctiveness, strength) Phenomenal consciousness C1:Consciousness involves two dimensions: Subjective experience and control • Availability to phenomenal consciousness depends on both the potency of representations and by their availability to control Banbury — 21

  22. implicit cognition priming explicit cognition automaticity Availability to P-Consciousness Potency Availability to A-Consciousness Availability to Control QUALITY OF REPRESENTATION (stability, distinctiveness, strength) Phenomenal consciousness C2: Availability to consciousness correlates with quality of representation • implicit representations are not available to conscious experience • Explicit representations constitute the dominant focus of consciousness • automatic representations constitute the periphery of consciousness C5:Learning shapes conscious experience Banbury — 22

  23. implicit cognition priming explicit cognition automaticity Availability to P-Consciousness Causal Efficacy Availability to A-Consciousness Availability to Control QUALITY OF REPRESENTATION (stability, distinctiveness, strength) Ways to be implicit • Because weak representations are involved: unconscious learning and priming; implicit knowledge as knowledge without consciousness • Because stronger, conscious representations are not accompanied by relevant metaknowledge: implicit learning as the indirect effects of explicit learning; implicit knowledge as conscious knowledge without metaknowledge • Because “automatic” representations can not be controlled: automatic uses of memory; implicit knowledge as conscious knowledge without control Banbury — 23

  24. A cognitive hierarchy • The brain’s functional and anatomical organization involves many interconnected networks sensitive to increasingly abstract dimensions of the stimulus (from PC to HC and FC) • Skill acquisition and learning involves both “moving up” in this hierarchy as well as changes within the modules • The dominant contents of consciousness consist of the networks in which most change is currently taking place (those that require the most control) Banbury — 24

  25. >>> Being Virtual • The framework suggests the necessary conditions under which representations are most likely to be available to form the contents of conscious experience • What is missing? What might the sufficiency conditions be? • Importance of self-representations acquired through processes of learning Banbury — 25

  26. Assumptions about Self S1: A crucial adaptive advantage for any organism is its ability to predict future states of its environment S2: Successful anticipation of future states in an environment that changes constantly requires organisms equipped with learning mechanisms S3: Successful anticipation of future states based on current states requires a model of the environment to be built. Such a model can be extremely simple, consisting of elementary associative links between current states and future states, or very complex, consisting for instance of a simulation of relevant aspects of the environment such that the future consequences of current actions can be explored in a flexible way Banbury — 26

  27. Assumptions about Self S4: When the environment includes other agents, and particularly potentially hostile agents, a crucial adaptive advantage for any organism is its ability to successfully predict the behavior of these agents S5: Successful anticipation of the behavior of other agents requires a model of how the behavior of these agents is influenced by the environment and by their own internal states. Again, such models can be very simple or very complex. More complex models, because they are more flexible and more detailed, provide adaptive advantages to the organism that possesses them Banbury — 27

  28. Assumptions about Self S6: From 1—5, it follows that organisms equipped with sufficiently powerful learning mechanisms and with sufficiently developed neural resources will develop detailed models of the internal states of agents it encounters in its environment S7: An organism that has developed such a detailed model of other agents is conscious in the fullest sense because in so doing it has developed the ability to entertain a third-person perspective on itself. This third-person perspective of a system upon itself arises when this system has developed a simulation of itself based on its simulation of other agents: You are a simulation Banbury — 28

  29. Conscious machines? • Minimal conditions for consciousness: • Massive information-processing resources that are sufficiently powerful to simulate certain aspects of their own inner workings. • A continuously learning system that attempts to predict future states • Immersion in a suitably rich environment from which models of yourself can be built Banbury — 29

  30. Conclusions • Consciousness is a graded, dynamic phenomenon that involves both a continuum as well as a dichotomy • Learning is a fundamental aspect of consciousness: • Learning shapes conscious experience • Conscious experience shapes learning • Learning is mandatory, and has both direct and indirect effects • The function of consciousness is to offer flexible, adaptive control over behavior • Your “self” is a simulation of other minds • Machine consciousness is possible Banbury — 30

More Related