The Personality Insights service offers a set of core analytics for discovering actionable insights about people and entities. The following sections provide basic information about using the service.
The Personality Insights service is based on the psychology of language in combination with data analytics algorithms. The service analyzes the content that you send and returns a personality profile for the author of the input. The service infers personality characteristics based on three models:
Big Five personality characteristics represent the most widely used model for generally describing how a person engages with the world. The model includes five primary dimensions:
Each of these top-level dimensions has six facets that further characterize an individual according to the dimension.
Needs describe which aspects of a product will resonate with a person. The model includes twelve characteristic needs: Excitement, Harmony, Curiosity, Ideal, Closeness, Self-expression, Liberty, Love, Practicality, Stability, Challenge, and Structure.
Values describe motivating factors that influence a person's decision making. The model includes five values: Self-transcendence / Helping others, Conservation / Tradition, Hedonism / Taking pleasure in life, Self-enhancement / Achieving success, and Open to change / Excitement.
For detailed information about the personality models and their characteristics, see Personality models.
The service can also return an indication of the consumption preferences for the author of the input text. Based on the personality characteristics inferred from the input, consumption preferences indicate the author's likelihood to pursue different products, services, and activities. The service assigns a binary or ternary indication of the author's inclination for each preference.
The service groups the individual preferences into eight categories: Shopping, Music, Movies, Reading and learning, Health and activity, Volunteering, Environmental concern, and Entrepreneurship. Each category contains from one to as many as 13 individual preferences.
For more information about the consumption preferences, their meaning, and their range of possible values, see Consumption preferences.
You can submit input as plain text, HTML, or JSON content, and the service can return its analysis in JSON or CSV format. For each personality characteristic, the service reports a percentile, which is a normalized score that describes the extent to which the author's writing exhibits a characteristic within a sample population. If requested, the service also returns a raw score, which is an absolute value that is not normalized for a sample population.
If the input is timestamped, the service provides a summary of the author's writing habits with respect to day of week and time of day. If requested, the service also returns a likelihood score for each of its available consumption preferences. For more information about submitting a request and about the meaning of the different results, see Input: Requesting a profile and Interpreting the numeric results.
A meaningful personality profile can be created only where sufficient data of suitable quantity and quality is provided. Because language use varies naturally from document to document and from time to time, a small sample of text might not be representative of an individual's overall language patterns. Moreover, different characteristics and different media converge at somewhat different rates.
Up to a point, more words are likely to produce better results, improving the service's precision by reducing the deviation between the predicted results and the author's actual score. You can send the service up to 20 MB of input content, but accuracy levels off at around 3000 words of input; additional content does not contribute further to the accuracy of the profile. Therefore, the service extracts and uses only the first 250 KB of content, not counting any HTML or JSON markup, from large requests.
This figure does not map to an exact number of words, which varies based on the language and nature of the text. In English, for example, average word length is between four and five characters, so this figure provides around 50,000 words, which is at least 15 times more words than the service needs to produce an accurate profile. By truncating long input, the service improves response time without sacrificing precision. The
word_count field of the response JSON indicates the number of words that the service actually uses for a request, which can be less than the number of words submitted.
The following table documents the average Mean Absolute Error (MAE) across all characteristics based on the number of words provided as input. The smaller the MAE, the closer the service's results are to the scores the author would receive by taking an actual personality test. The final column reports the average correlation between inferred and actual scores across all characteristics. The information is based on English-language data, but the general guidelines for the number of words to provide apply to all languages. For more information about average MAE and correlation, along with statistics for specific languages, see How precise is the service.
|Number of words||Average MAE
across all characteristics
across all characteristics
IBM recommends that you provide at least 1200 words of input text, which results in an MAE that is within two percent of the best MAE the service can return. Submitting between 600 and 1200 words results in an MAE that is within three percent of the best MAE that IBM still considers acceptable. Providing 3000 words is sufficient to achieve the service's maximum precision. You must submit at least 100 words; otherwise, the service reports an error. If you submit fewer than 600 words, the service reports a warning but still analyzes the input text.
These guidelines can help you accommodate the reliability of the results to your application. For example, for a casual application that recommends movies, you might be comfortable with less precision; for an application that makes more critical recommendations, you likely require more precise results. For more information about how the service infers personality characteristics, see How personality characteristics are inferred.
The Personality Insights service can infer characteristics and return its response in the following languages.
|Request languages||Response languages|
You can use any combination of supported languages for the request and response, but all input text must be in the same language. For more information, see Specifying request and response languages.
IBM researchers and others have validated the hypotheses behind the Personality Insights service via studies in multiple domains. In addition to influencing the direction of the service and ensuring the validity of its results, these studies offer ideas for applying the service in your applications. For information about use cases that suggest additional applications of the service, see Use cases.
The following table lists studies performed by IBM researchers and briefly summarizes their findings. These studies contributed directly to the development of the service. For more information about these IBM studies, see Studies by IBM researchers.
|Study||Subjects||Effect of Personality Characteristics|
Responding to tweets
|2000 Twitter users||
More likely to respond: Score high on Big Five dimensions
and facets excitement-seeking, friendliness, activity level,
gregariousness, trust, morality, extraversion, and
Less likely to respond: Score high on Big Five facets cautiousness and anxiety
|3500 Twitter users||More likely to re-tweet: Score high on Big Five dimensions and facets modesty, openness, and friendliness|
|More than 6000 Twitter users||Respond more favorably: Score high on Big Five dimension openness, and score low on Big Five dimension emotional range|
|Thousands of retail customers||More likely to respond: Score high on Big Five facets orderliness, self-discipline, and cautiousness, and score low on Big Five facet immoderation|
|600 Twitter users||Predictors of brand preference: Big Five dimensions and facets conscientiousness, conservation, self-enhancement, and agreeableness; Needs love and ideal; and Value hedonism|
|More than 3000 community members||Predictors of member satisfaction: Characteristics such as anger, anxiety, work, leisure, inhibition, assent, and use of the first-person pronoun|
|More than 200 participants||
Interest in environmental articles: Score high on Value
Interest in work-related articles: Score high on Value self-enhancement
|Hundreds of thousands of data records||Predicting consumer preferences: Demographic data and personality characteristics|
The following table lists and briefly describes studies by researchers outside of IBM. These studies influenced the direction of the service, and they validate the hypotheses on which it is based. For more information about these external studies, see Studies by other researchers.
|Consumer preferences||Marketing messages||People respond more favorably to marketing messages tailored to their personality.|
|Online marketing||People with high openness are more likely to try new things.|
|Car ownership||People who own convertibles differ in personality from the owners of standard or compact cars.|
|Car ownership||People who own powerful cars differ in personality from those who prefer not to own powerful cars.|
|Car ownership||Car owners tend to perceive the types of cars they buy as an extension of their personality.|
|Product design||People with high openness respond to product design, while people with low openness value other product traits.|
|Brand preference||Personality characteristics affect brand preference.|
|Personal preferences||Music preference||People respond more positively to music recommendations tailored to their personality.|
|Music appreciation||Individual differences in personality and cognition might determine how music is experienced.|
|Movie preference||Movie preference correlates with personality.|
|Spending habits||Credit card use||Individuals with high conscientiousness tend not to abuse credits cards, while those with high emotional range do.|
|Risk profiles||Financial investment||Personality influences risky decision-making in financial investments.|
|Taking risks||Agreeableness and conscientiousness lower people's willingness to take risks, while extraversion increases it.|
|Adventure sports||Personality characteristics directly influence the probability of participating in risky sports.|
|Skydiving||Skydivers typically have higher excitement-seeking and adventurousness than the general population.|
|Professional performance||Job performance||Conscientiousness and other Big Five dimensions affect job performance.|
|Job performance||Conscientiousness is the characteristic that best predicts job performance.|
|Leadership||Extraversion correlates with leadership abilities.|
|Job satisfaction||Personality characteristics such as extraversion and introversion positively and negatively indicate job satisfaction.|
|Adaptability||Conscientiousness, extraversion, and openness correlate with career adaptability.|
|Academic performance||Academic achievement||Personality characteristics such as neuroticism and conscientiousness can negatively and positively indicate academic performance.|
|Self-improvement||Openness and conscientiousness often lead to more self-improvement learning, while neuroticism decreases such motivation.|
|Professional relationships||Interaction preferences||Conscientiousness, openness, and other Big Five dimensions affect patients' interaction preferences with medical professionals.|
|Patient satisfaction||Openness and conscientiousness affect physicians' patient satisfaction.|
|Personal relationships||Romantic relationships||Differences in personality characteristics such as agreeableness, emotional range, and openness predict marital dissatisfaction.|
|Health||Health-related outcomes||Personality characteristics predict health-related outcomes such as self-rated physical health and absenteeism from work.|
|Longevity||High scores in conscientiousness, extraversion, and openness are associated with greater life expectancy.|
|Life outcomes||Personality characteristics predict life outcomes such as mortality, divorce, and professional success.|
|Diet and exercise||High-fat food||People with high scores in emotional range and especially in immoderation tend to consume high-fat food.|
|Low-fat food||Conscientiousness and especially self-discipline and dutifulness influence the consumption of low-fat food and regular weight control.|
|Healthy food||The choice to consume healthy food is related to personality characteristics.|
|Varied diet||Food lovers score high in openness to experience, which might motivate them to try different foods.|
|Healthy lifestyle||Personality influences the decision to pursue a healthy lifestyle, including physical exercise.|
|Dining out||Dining choices||Openness to experience correlates positively with more frequent dining out.|
|Dining choices||Excitement-seeking also correlates with the decision to dine out.|
|Environmental consciousness||Environmental concerns||Personality influences people's level of interest in environmental concerns and pro-environmental behavior.|
|Product decisions||Personality influences the decision to buy environmentally conscious products such as low-emission vehicles.|
|Community service||Community engagement||Personality affects the tendency for prosocial behavior such as community service.|
|Religion and spirituality||Religious tendencies||Agreeableness suggests greater religious tendencies, while authority-challenging indicates less motivation to participate in religious and spiritual practices.|