Category Archives: Uncategorized

Psychology: Introduction and Future Research Directions

Update 2: This is outdated. The better framework is a quantitative one.

Update: The following represents initial research on the topic. I have a new framework which I may present this summer.

Cognitive biases are patterns of thought which cause mistakes. A simpler way of putting it is that cognitive biases are modalities of human thought, since immoderate use of any modality is an imprudent mistake. Modalities include reward response, influence from social proof, and influence by association, which in the language of cognitive biases put forward by Charlie Munger are reward super-response, over-influence from social proof, and influence by mere association, respectively. Perspicacious writers like Scott Adams claim cognitive biases are central to their worldview. So there must be something to the idea.

Another interesting idea is psychological typing, which is the classification of people based on their individual psychology. When quantitative, this is called psychometrics. Advertising companies use psychological typing combined with cognitive biases to influence people more effectively. The idea is that different types of people respond to different modalities, so the modalities in an ad should be tailored to the person viewing the ad. Of course, the utility of typing doesn’t end there. Understanding people better lets you manage them better, find better roles for them, and diagnose their problems more effectively.

Different typing systems are interesting to learn about. The most scientific and easy-to-understand one is Big Five. The Big Five model is based on the lexical hypothesis, which is the assumption that personality can be accurately constructed by considering the adjectives people use to describe themselves and others. Motivated by this hypothesis, late 20th century scientists like Lewis Goldberg took a large number of personality surveys and repeatedly performed factor analysis, until they obtained five factors of personality. These are the Big Five. The Big Five traits have been validated cross-culturally and have been shown to have a genetic component. Psychologists such as Geoffrey Miller hold that the Big Five model is by far the best way to do psychological typing.

The logical hole in lexical analysis is that average people may not be perceptive or articulate enough to fully describe personality. So an alternative approach is to rely on genius. That’s where Carl Jung comes in. Jung spent a few years in a cabin before publishing Psychological Types in 1921. The book defines eight types and provides lengthy analyses of each one.

Are Jung’s types useful? Maybe. [Edit: After going over Andrew Ng’s machine learning course, I revise this to “slightly.”] His ideas were certainly persuasive. 89 out of the Fortune 100 companies use his types, slightly revised and rebranded as the Myers Briggs Type Indicator (MBTI). Some users of MBTI seem to take it a little too far, categorizing themselves and their friends using the system. Carl Jung must have been well-aware of MBTI when near his death in 1961, he wrote in Man and His Symbols that there was “nothing dogmatic” about his types, though they are “useful [for] understanding one’s own prejudices.”

In summary, there are several personality typing systems. There’s Big Five, which is scientific. There are carefully thought-out constructs like Jung’s Types and MBTI. Finally there are of course ad-hoc, domain-specific typing systems. For example, in the context of business, Steve Jobs talks about sales people vs. product people and process people vs. content people.

Our goal is not a survey of personality typing systems, but the acquisition of useful models of human nature. A natural question now is the best way to connect cognitive biases to psychometrics. In other words, what cognitive biases are most typical for different personality types?

One way to cheat is to look at a list of cognitive biases, such as the one in Charlie Munger’s speech The Psychology of Human Misjudgment, and define personality as one’s susceptibility to each bias. But then you’re throwing away a lot of insight, given the large amount of existing research on personality psychology.

A better approach is to take a list cognitive biases and try to merge it with state-of-the-art personality psychology frameworks. Though I’m not the first to have this idea, this produces many insights which I believe are novel.

For example, one cognitive modality is incentives. The way people are paid or rewarded can have large effects on their behavior and beliefs. Incentive-reward is linked to the introversion/extroversion factor of personality. Extroverts happily do a large number of things in response to reward, while introverts tend to resist temptations and prefer doing a smaller number of things. Given this observation, introverts and extroverts might consider setting goals in different ways. Introverts might focus on setting deeply rewarding goals, while extroverts might focus on avoiding unproductive or unethical tasks. Fitting behavioral changes to temperament in this way may make them more effective.

Another cognitive modality is consistency. Things people learn stick in their heads, closing them to alternative perspectives. Conscientious people have a tendency to enforce internal and external boundaries, which helps explain the correlation between conscientiousness and political conservatism. So conscientious, low-openness people might at first be less affected by consistency bias, since they avoid learning suspicious ideas which could bias them in the future, such as questionable political ideas. Put differently, they have stronger mental immune systems. But if they do end up learning something, the consistency bias might affect them more, since they’ll be less curious about alternative hypotheses.

Finally, excessive self-regard is a cognitive bias where people overestimate themselves and their possessions. We might think this is related to narcissism, a Dark Triad trait related to the extroversion and disagreeableness factors of the Big Five. But it may also be related inversely to neuroticism, which is associated with a tendency to fear and anxiety. Low social status and a lack of serotonin correlate with neuroticism. A less neurotic and more emotionally stable person may be more prone to overestimating themselves.

Analyzing the relationships between psychometrics and cognitive biases produces interesting directions for new research. Well-studied personality constructs have a lot of interesting philosophy attached to them, and cognitive biases themselves are important tools in marketing, investing, and decision-making. Keeping track of both in the same context yields interesting insights.


Personality typing systems:

  • Principles by Ray Dalio. Chapter II.4. Previous version (dead link):
  • Psychological Types by Carl Jung. Originally published in 1921. Available in Collected Works of C.G. Jung, published in 2014.
  • The Big Five is thoroughly described in course notes and publications by Jordan Peterson.
  • HEXACO, which is similar to The Big Five, has better cross-cultural validation. For summaries of HEXACO and others, see
  • MBTI is described in Building Blocks of Personality Type by Leona Haas and Mark Hunzicker.

Cognitive bias systems: