Chapter 2: Personality Assessment

The imp of personality assessment

Prevalent/common:

• Coming in many shapes & forms

• Some scientific & validated, many not

Impacts far beyond academia: Companies — hiring, retention, & promo decisions

Expensive: It’s a billion dollar industry

How to measure personality

It depends on who we think about personality! Examples:

1. Trait/humanistic approach: people have clear traits / goals that define who they are → we can just ask

2. Psychodynamic approach: people's behaviours are driven by unconscious processes → can't just ask!

Projective Tests

• Important thoughts, feelings, and motives operate outside of conscious awareness

• Projective hypothesis: If a person is asked to describe or interpret ambiguous stimuli their responses will be influenced by nonconscious needs, feelings, and experiences

• In interpreting such a stimulus people project their unique personality onto the stimulus

Rorschach Inkblots & TATs

Two of the most famous projective personality tests:

• Rorschach inkblots: Individuals are asked to interpret symmetrical blots of ink

• TAT: Individuals are asked to tell a story based on a series of ambiguous pictures

• Idea: whatever they see is not actually on the card / in the pictures, but reflects the contents of their mind

Projective tests: Prevalence

• Still frequently used (very common in therapy, court, etc.) even though scholars caution against it

• 4th most used test in psyc

Projective tests (Rorschach & TAT) Pros

• Ice breaker to get clients to open up

• Rorschach: Some utility for prediction of outcomes (i.e., suicide, attendance at treatment sessions, commitment to a mental hospital)

• TAT: some evidence for assessment of implicit motives

Projective tests (Rorschach & TAT) Cons

• Scarce validity evidence

• Expensive and time-consuming

• No objectivity: unclear what they really mean

• Other, less expensive tests work as well or better

Objective Tests

• Most familiar and most widely used approach to personality assessment

• Idea: personality tests that are more objective and less open to interpretation

• Objective personality test = test consisting of a list of questions to be answered using a limited set of response options (e.g., true or false; strongly disagree, slightly disagree, slightly agree, strongly agree)

• Responses to these items then are scored in a standardized, predetermined way

  • Illustration: self-ratings on items assessing talkativeness, assertiveness, sociability, adventurousness, and energy can be summed up to create an overall score on the personality trait of extraversion

Varieties of Objective Tests (and pros)

• Type S, I, or B data

• Comprehensiveness: unidimensional (e.g., Rosenberg Self-Esteem Scale) or multidimensional (e.g., Big Five Inventory 2)

• Length: Short vs full scale

  • Principle of aggregation: Averaging answers to multiple items decreases error and increases stability and reliability 

• Breadth of target characteristics: Big Five dimensions versus facets versus nuances

Principle of aggregation

Averaging answers to multiple items decreases error and increases stability and reliability

Constructing objective personality: Rational method

Write items that seem directly, obviously, and rationally related to what is to be measured

• Generates S-data

• Items are developed based on theory and expert knowledge

• Ideally informed by elaborate theory on the concept that is to be measured; in practice this is sometimes less systematic

2 Steps of rational method(example: wanderlust)

1. definition and specification of the construct (what behaviours, experiences, emotions etc. are manifestations of wanderlust?)

2. item generation (collect indicators that can be assessed through statements or questions; example: “I like to discover new cities.”; “When I am away from home, I like to explore the local vibe.”

4 conditions for validity for rationally constructed pers tests to work

1. Items mean the same thing to the test taker and creator

2. Test taker is able to make an accurate self-assessent (cognitively able)

3. Test taker is willing to make an accurate and undistorted report

4. The items are valid indicators of what the test is trying to measure

Even tho most rationally constructed pers tests do not meet all 4 conditions, this remains the most common form of test construction

Constructing objective personality: Factor analytic method

identifies groups of items that have something in common (factor: the property that ties the together)

Used to decide how many fundamental traits exist (e.g. Big Five: Openness, conscien., etc)

4 steps to apply factor analysis

1. Generate a long list of objective items

2. Administer these items to a large number of ppl

3. Analyze w a factor analysis

4. Consider what the items that group tgt have in common and name the factor

Constructing objective personality: Empirical method

Select items that distinguish most clearly between pre-determined groups

5 steps of empirical method

1. Gather lots of items

2. Have a sample of ppl alr divided into groups (e.g. based on occupation or diagnostic criteria)

3. Run the test

4. Compare the answers of the diff groups, select the items that are most useful to differentiate the diff groups

5. Cross-validation: Check whether the selected items distinguish equally well in a diff sample

Empirical construction method example

• MMPI-2; used for clinical assessment

• Compare items between two groups (psychiatric diagnosis vs no diagnosis)

• Based on statistical parameters, selection of the 556 items that best distinguish between the two groups

Best use for each test (Rational, factor analytic, empirical)

Rationally constructed tests: Best when you want a clear, theory-based measure of a specific construct (e.g., BDI for depression)

Factor-analytic tests: Best when you want to discover/confirm underlying dimensions of personality or traits. (Big five). 

• Statistical correlations.

• Identifies clusters of things that have something in common

Empirical tests: Best when you need a test that predicts real-world group differences (e.g., clinical vs. non-clinical) (e.g. MMPI)

Digital Footprints as Personality Measures – Procedure

• volunteers complete a classic self-report personality questionnaire to assess the Big Five and provide sociodemographic characteristics.

• Researchers gain access to the individuals' personal Facebook profiles and collect data on Facebook likes (with consent).

• A machine learning algorithm learns which likes are associated with which personality traits.

The machine learning algorithm can now apply this knowledge to new Facebook profiles and predict users’ personality traits based on their publicly accessible likes.

How Accurate Are These Measures? (Assessing personality from activities on Facebook)

• Half as accurate than using typical test-retest reliability for survey-based self-report measure, especially the Big Five, Satisfaction with Life, Intelligence (Proved that the generic S data is much accurate)

• Openness is the only component that shows similar (r) in Facebook Likes and traditional questionnaire

Digital footprint’s prevalence

S-data assess big 5, sociodemographic characteristics

Access & collect data from personal facebook profile & likes

Machine learning algorithm learns association between likes & personality traits

Algorithm applies knowledge to new profiles & predicts personality based on likes

Personality can be reliably inferred on social media from…

• Facebook likes

• Flickr images

• GPS data

• Instagram captions

• Music preferences on spotify

• Smartphone usage patterns

• Spending records

• Twitter usage

Common pitfalls of studying personality thru social media activity

• Models can be very culturally specific or get outdated and predictions may fail

• Ground-truth problem: Is alignment w self-report necessarily the best metric for accurate personality assessment? (social media might not be your true self)

• Ethics and privacy concerns surrounding data sourcing

Can Language Models judge public figures’ personality?

Chat-GPT’s ratings were highly and consistently correlated w the avg human ratings