Collective Predictive Coding Hypothesis for Collective Curiosity and Exploration

Transcript

1. Association for Philosophy of Science 27th Oct 2024 Collective Predictive Coding Hypothesis for Collective Curiosity and Exploration Hiro Taiyo Hamada, Ph.D 1) Research Team Lead, Araya Inc. 2) Committee Chair, DeSci Tokyo @HiroTHamadaJP
2. Curiosity as Personal Driver of Scientific Activities
3. How does this personal driver lead to global scientific progress?
4. Contents Collective Predictive Coding as Models of Science (CPC-MS) 1. Looking back CPC-MS from individual selections 2. Theory of Individual Curiosity 3. Collective Curiosity Beyond Individual Curiosity 4. Implications 5.
5. Collective Predictive Coding as Models of Science (CPC-MS) Global Scientific Representations Agents Scientist World Target System Target d o d i Data Observation Experiment Internal representation z d i External representation w d [Taniguchi et al., arxiv, 2024]
6. Looking back CPC as individual selections Global Scientific Representations Agents Scientist World Target System Target d o d i Data Observation Experiment Internal representation z d i External representation w d
7. Looking back CPC as individual selections Global Scientific Representations Agents Scientist World Target System Target d o d i Data Observation Experiment Internal representation z d i External representation w d Theme Selection m
8. Looking back CPC as individual selections Global Scientific Representations Agents Scientist World Target System Target p( | , ) o d i Data Observation Experiment Internal representation z d i External representation w d Experiments a d k Hypothesis Updating z d i w d a d k
9. Global Scientific Representations Agents Scientist World Target System Target p( | , ) o d i Data Observation Experiment Internal representation z d i External representation w d Experiments a d k Hypothesis Updating z d i w d a d k What drives actions/experiments/exploration for theme selection and hypothesis testing? Looking back CPC as individual selections
10. What drives actions/experiments/exploration for theme selection and hypothesis testing?
11. Daniel Berlyne 1924-1976 Information Gain as Epistemic curiosity Curiosity: An internal driver for exploratory behaviors [Berlyne, 1954, 1960, 1966]
12. Daniel Berlyne 1924-1976 Information Gain as Epistemic curiosity Perceptual curiosity: The curiosity which leads to in-creased perception of stimuli. [Berlyne, 1954, 1960, 1966] Epistemic curiosity: The curiosity is a “drive to know”. Reduction of uncertainties and acquisition of knowledge.
13. Information Gain as Epistemic curiosity [Lowenstein, 1994] knowledge (uncertainty) Curiosity low high
14. Deprivation Sensitivity is associated with concept clustering [Lydon-Staley et al., Nat Hum Behav, 2021] N = 149, 5 hours/day, 21 days
15. Deprivation Sensitivity is associated with concept clustering [Lydon-Staley et al., Nat Hum Behav, 2021] N = 149, 5 hours/day, 21 days Individual curiosity is associated with connections with known knowledges.
16. Concept network growth is associated with impact of science. [Kedrick et al., Nat Hum Behav, 2024] [Ju et al., arxiv, 2020]
17. Concept network growth is associated with impact of science. [Kedrick et al., Nat Hum Behav, 2024] [Ju et al., arxiv, 2020] Concept network growth ~ collective exploration driven by collective curiosity
18. Collective curiosity toward trends vs. against trends Toward trends Aims core in the network but may be just periphery.... Avoid trends in the network but can find new connections.... Against Trends Your proposal Field A Field A Field B
19. Collective curiosity beyond individual Curiosity Global Scientific Representations Agents Scientist World Target System Target d o d i Data Observation Experiment Curiosity z d i Concept Network w d Influence
20. How does this personal driver lead to global scientific progress?
21. How does this personal driver lead to global scientific progress? Individual curiosity is constrained and driven by concept network/global representation. 1. Dynamic structural change of concept networks recursively by individual curiosity 2.
22. Conclusions & Implications Collective curiosity is associated with information gaps in concept networks. 1. Scientific concept network is a global representation in the CPC-MS framework. 2. In the CPC-MS framework, individual curiosity is affected by global representations (which also drives collective curiosity). 3. From concept networks, we including AI scientists may predict/generate emerging fields of science. 4.
23. global representation from scientist/Industry community A global representation from AI scientist global representation from citizen/scientist community B Diverse “Global Representation” including AI and citizen scientists may generate totally further new concept network growth by collective curiosity. CPC can be a model for how multi types of science can be aka meta-science.