search

Week 2: Epistemics

chevron-rightReasoning transparencyarrow-up-right (30 minutes)hashtag

  • Benefits of Reasoning Transparency

    1. Fciliate the process of updating other's view in response

  • How?

    1. Open with a linked summary of key takeaways

    2. Indicate primary & secondary considerations for your key takeaways

    3. Indicate confidence levels

      1. 0-100%, plausible, seems

    4. Inform about concluding variables underlying confidence levels

      1. A reasoning that seems robust

      2. shallow/careful examination of one or more studies that you feel weakly/strongly qualified to assess

      3. Verfiable facts coua can provide sources for

      4. Expert opinions

      5. Vague impression

      6. e.g.,

        1. my view is based on a large number of undocumented conversations, such that I don’t think it is realistic to aim for being highly convincing in this post.

        2. It is widely believed, and seems likely, that regular, high-quality sleep is important for personal performance

        3. [CBT-I] can be delivered on an individual basis or in a group setting,

        4. As far as we know, the vast majority of human cognitive processing is unconscious, including a large amount of fairly complex, ‘sophisticated’ processing. This suggests...

    5. Inform about Crucial Considerations

    6. Inform experience, trustworthiness and fairness

chevron-rightAaron’s epistemic storiesarrow-up-right (15 minutes)hashtag

  • Finding Trust-worthy sources of information

    • Fact-Checking sources of information e.g., news

    • Does this journalist/news update their views?

  • Attune to optimism & pessimism bias

    • Check on 'Predictions' e.g., 'next big thing', 'time-lines'

    • Align goals with reality e.g., tasks & completion time

chevron-rightHow do you predict the future? Ask samotsvetyarrow-up-right (15 minutes)hashtag

  • Benefits

    • Better Predictions = better decisions = better actions = better impact

    • Understanding the model of the world

  • How?

    • Thinking in numbers > words e.g. 'unlikely'

    • Two Key Steps

      • Acknowledge that the future & past will look mostly the same

      • Isolate the small cases where the two are different

    • Practice develops the skill of forecasting

    • Updating one's forecasts

    • Base rate

      • rate at which some event tends to happen

      • e.g., NY Yankees won 27 times the world serious out of 119 i.e., 22.7% is the base rate

  • Difficulty predicting the cases of outliers

    • e.g., nuclear bomb hitting London

chevron-rightScott Alexander, Beware the Man of One Studyarrow-up-right (15 minutes)hashtag

  • Evidence is complex

    • Due to the nature of experiments there will usually be a bell curve i.e., opposing results while finding middle ground

  • Decrease your confidence about most things if you're not sure that you've investigated every piece of evidence

  • Do not trust biased sources

    • e.g., websites

  • When receiving 'overwhelming evidence' do a simple search to see if the opposite side has equally 'overwhelming evidence'

chevron-rightMichael Aird on ‘independent impressions’ vs. ‘all-things-considered beliefs’arrow-up-right (5 minutes)hashtag

  • Independent Impressions

    • What you believe about X if you do not consider updating your beliefs despite receiving peer opinions

  • All-things-considered beliefs

    • Takes into account peer opinions

  • Suggestions

    • Recognize & distinguish independent impressions from all-things-considered beliefs

    • State clearly about expressing independent impressions or a-t-c beliefs

    • Feel comfortable expressing our independent impressions

    • Make decisions based on a-c-t beliefs > independent impressions

chevron-rightSequence thinking versus cluster thinkingarrow-up-right (45 minutes)hashtag

Sequence Thinking

  • What?

    • Decisions based on a single model of the world

    • Belief 1 -> Belief 2 -> 3 -> 4 -> Conclusion

    • If each step A, B, C, ...,N holds true, then the conclusion X logically follows

  • Benefit

    • Generate (novel) ideas

      • e.g., contradict conventional wisdom

    • Deep Understanding

    • Transparency

      • Each assumption is explicit

      • Causal Understanding

  • Disadvantage

    • One big factor (even if uncertain) can domiante hte final conclusion

    • Vulnerability to Errors

      • If any single assumption is flawed, the entire conclusion might be flawed

      • Overconfidence/Underconfidence is not cushioned i.e., will have a strong impact on one's conclusion

      • Facilitates 'end justifies the means'

      • Facilitates non-exploration of crucial considerations

Cluster Thinking

  • What?

    • Decisions based on multiple models of the world

    • Weighing of all: Perspective 1 implies X, Perspective 2 implies Y, Perspective 3 Implies X

    • “I think we should do X, but I can’t say exactly why, and some of the most likely positive outcomes of this action may be outcomes I haven’t explicitly thought of.”

    • Cushions high weights

      • De-weighing of single beliefs/models even if high expected value

  • Benefit

    • Making final decisions as it is more balanced

    • Robustness

      • Avoids the potential wrongness of a single-model world

  • Disadvantage

    • Complexity

      • May lack clarify of a single-model approach

      • Weighing & integrating various perspectives might be challenging

    • Conflict

      • Diverse view points might lead to conflicting conclusions

How To Think?

  • Robustness & uncertainty

    • probability that my view remains stable despite gaining more information

  • Regression to normality

    • Unusual claims require strong evidence to make a shift

  • How to Think?

    • Combination of Sequence & Cluster Thinking

      • Sequence to explore possibilities & cluster to evaluate them comprehensively

    • Well-understood situation: sequence > cluster

    • Complex / less understood situation: cluster > sequence

    • Recognize advantages / disadvantages of each

Personal Reflections

  • World as a cluster

    • Every individual uses sequence thinking i.e., 'belief 1 -> believe 2' even if that includes cluster thinking i.e., they come out with one conclusion

    • As a world, we share our sequence thinking conclusions and come to harmony i.e., cluster thinking

chevron-right Accurately predicting the future is central to absolutely everything. Phil Tetlock has spent 40 years studying how to do it better.arrow-up-right (2 hours)hashtag

Benefits of Forecasting

  • Probabilities guide our decisions and our behavior

    • Lack of probabilities hence may lead to mistakes e.g., 'impossible', 'possible', 'certain'

  • Superforecasters demonstrate the ability to make reliable forecasts even for 'unique' events

    • e.g., Brexit, Russia-Ukraine

  • World would be better off if government leaders would consult technocratically run forecasting tournaments for estimates on key issues

Hindrances to Forecasting with Probabilities

  • In a polarized world, one makes the claim that one is only e.g., 75% convinced of one side - one could face detrimental consequences

Caution

  • Counterfactual Reasoning is ideologically self-serving

    • Conservatives

      • Without President Reagan, Cold War would have continued and Soviets would've pushed further

    • Liberals

      • Soviet was economically collapsing and Reagan wasted billions in unnecessary defense expenditures

  • Claims imply counterfactual claims

    • e.g., Every time you claim someone to be a 'good' or 'bad' president you imply the claim about hwo the world would have unfolded in an alternative universe to which you have no empricial access

How?

  • Willingness to Update

  • Probabilistic Thinking

    • Discriminate between differences as fine as 56% versus 57%

  • Macro into Micro

  • Acknowledge Confidence (Over/Under)

PreviousWeek 1: Opportunity of Effective AltruismNextWeek 3: AI Safety and Security

Last updated