warmthcompetence Vignette

Warmth and Competence

Warmth and competence are the two main dimensions of social perception and judgment (Cuddy et al. 2008) . When individuals introduce or describe themselves, their audiences automatically make judgments about their warmth and competence. In the warmthcompetence package, we provide tools that estimate warmth and competence social perceptions from natural self-presentational language. We use pre-trained enet regression models to provide numerical representations of warmth and competence perceptions.

Installation

To install warmthcompetence from CRAN, use the following code in your R session:

install.packages("warmthcompetence")

To install the development version from GitHub, use the following code:

devtools::install_github("bushraguenoun/warmthcompetence")

First, we load the package into our environment.

library("warmthcompetence")

Note that some features depend on spacyr which must be installed separately through Python. To install spacyr, follow the instructions here. Then run the code below:

spacyr::spacy_initialize()

Usage

warmthcompetence contains two main functions: warmth() and competence(). These functions can be used as described below:

competence_scores <- competence(text_vector, ID_vector, metrics = "scores")

warmth_scores <- warmth(text_vector, ID_vector, metrics = "scores")

In the code above, text_vector is the vector of texts that will be assessed for warmth or competence. ID_vector is a vector of IDs that will be used to identify the warmth or competence scores. The metrics argument allows users to decide what metrics to return. Users can return the warmth or competence scores (metrics = "scores"), the features that underlie the warmth or competence scores (metrics = "features"), or both the warmth or competence scores and the features (metrics = "all"). The default choice is to return the warmth or competence scores.

Example: Warmth and Competence Signaling in Self Introductions

We ran a study in which 393 participants were randomly assigned to either the warmth or competence condition and asked to write a short introduction of themselves that presented them as a [warm/competent] person. Then, we had three independent judges blind to participant condition rate each of the introductions for warmth and competence on a scale of 1 to 10.

The study data can be accessed by calling data("vignette_data"). This command will return a data frame with all of the data collected from the study. The columns of this data frame are the following:

ResponseId: participant identifiers
bio: participant introductions
condition: participant condition in the study (warmth versus competence).
RA_warm_AVG: the average perceptions of warmth for each participant introduction on a scale of 1 (extremely low warmth) to 10 (extremely high warmth) across three independent judges ( $\alpha = 0.76$ ). The individual RA ratings are in the following columns: warm_1, warm_2, and warm_3
RA_comp_AVG: the average perceptions of warmth for each participant introduction on a scale of 1 (extremely low competence) to 10 (extremely high competence) across three independent judges ( $\alpha = 0.79$ ). The individual RA ratings are in the following columns: comp_1, comp_2, and comp_3

Predicting Warmth Perceptions

Can our warmth model predict the average warmth ratings of the independent judges?

First we run the functions on the study data.

data("vignette_data")

# Generate competence scores
competence_scores <- competence(vignette_data$bio, vignette_data$ResponseId)

# Generate warmth scores
warmth_scores <- warmth(vignette_data$bio, vignette_data$ResponseId)

# Merge scores together
both_scores <- dplyr::inner_join(competence_scores, warmth_scores,
                                 by = c("ID" = "ID"))

# Merge scores into participant data
all_data <- dplyr::left_join(vignette_data, both_scores,
                             by = c("ResponseId" = "ID"))

Then, we explore whether the warmth model predicts the warmth ratings of the independent judges.

warmth_model <- lm(RA_warm_AVG ~ warmth_predictions,
                   data = all_data)

summary(warmth_model)
#> 
#> Call:
#> lm(formula = RA_warm_AVG ~ warmth_predictions, data = all_data)
#> 
#> Residuals:
#>     Min      1Q  Median      3Q     Max 
#> -3.6622 -0.2752  0.0374  0.3776  2.3164 
#> 
#> Coefficients:
#>                    Estimate Std. Error t value Pr(>|t|)    
#> (Intercept)         5.75999    0.03213  179.27   <2e-16 ***
#> warmth_predictions  0.79405    0.04660   17.04   <2e-16 ***
#> ---
#> Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
#> 
#> Residual standard error: 0.6369 on 391 degrees of freedom
#> Multiple R-squared:  0.4261, Adjusted R-squared:  0.4247 
#> F-statistic: 290.3 on 1 and 391 DF,  p-value: < 2.2e-16

As can be seen above, the predictions of the warmth model positively predict the independent judges’ ratings.

Predicting Competence Perceptions

Can our competence model predict the average competence ratings of the independent judges?

competence_model <- lm(RA_comp_AVG ~ competence_predictions,
                       data = all_data)

summary(competence_model)
#> 
#> Call:
#> lm(formula = RA_comp_AVG ~ competence_predictions, data = all_data)
#> 
#> Residuals:
#>     Min      1Q  Median      3Q     Max 
#> -3.9872 -0.4573 -0.0296  0.5009  2.9592 
#> 
#> Coefficients:
#>                        Estimate Std. Error t value Pr(>|t|)    
#> (Intercept)             5.48828    0.03785  144.98   <2e-16 ***
#> competence_predictions  0.99237    0.06899   14.38   <2e-16 ***
#> ---
#> Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
#> 
#> Residual standard error: 0.7502 on 391 degrees of freedom
#> Multiple R-squared:  0.346,  Adjusted R-squared:  0.3444 
#> F-statistic: 206.9 on 1 and 391 DF,  p-value: < 2.2e-16

As can be seen above, the predictions of the warmth model positively predict the independent judges’ ratings.

Predicting Participant Condition

Are model predictions of competence higher for the participants asked to present themselves as a competent person compared to a warm person?

t.test(competence_predictions ~ condition, data = all_data)
#> 
#>  Welch Two Sample t-test
#> 
#> data:  competence_predictions by condition
#> t = 5.9885, df = 389.68, p-value = 4.815e-09
#> alternative hypothesis: true difference in means between group competent and group warm is not equal to 0
#> 95 percent confidence interval:
#>  0.2132765 0.4217659
#> sample estimates:
#> mean in group competent      mean in group warm 
#>               0.1707312              -0.1467900

As can be seen above, the model significantly predicted that participants in the competence condition expressed more competence compared to those in the warmth condition.

Now, are model predictions of warmth higher for the participants asked to present themselves as a warm person compared to a competent person?

t.test(warmth_predictions ~ condition, data = all_data)
#> 
#>  Welch Two Sample t-test
#> 
#> data:  warmth_predictions by condition
#> t = -12.829, df = 372.43, p-value < 2.2e-16
#> alternative hypothesis: true difference in means between group competent and group warm is not equal to 0
#> 95 percent confidence interval:
#>  -0.8628778 -0.6335158
#> sample estimates:
#> mean in group competent      mean in group warm 
#>              -0.3610531               0.3871437

As can be seen above, the model significantly predicted that participants in the warmth condition expressed more warmth compared to those in the competence condition.

Conclusion

In this example, our pre-trained enet regression models predicted the ratings of independent human judges and participant judges. Feel free to use these models on your own data and reach out with any comments, questions, or concerns!

References

Cuddy, Amy J. C., Susan T. Fiske, and Peter Glick. 2008. “Warmth and Competence as Universal Dimensions of Social Perception: The Stereotype Content Model and the BIAS Map.” In Advances in Experimental Social Psychology, vol. 40. Elsevier. https://doi.org/10.1016/S0065-2601(07)00002-0.

Bushra Guenoun

2026-05-05