What's in a Name? Experimental Evidence of Gender Bias in Recommendation Letters Generated by ChatGPT.

BACKGROUND: Artificial intelligence chatbots such as ChatGPT (OpenAI) have garnered excitement about their potential for delegating writing tasks ordinarily performed by humans. Many of these tasks (eg, writing recommendation letters) have social and professional ramifications, making the potential social biases in ChatGPT's underlying language model a serious concern. OBJECTIVE: Three preregistered studies used the text analysis program Linguistic Inquiry and Word Count to investigate gender bias in recommendation letters written by ChatGPT in human-use sessions (N=1400 total letters). METHODS: We conducted analyses using 22 existing Linguistic Inquiry and Word Count dictionaries, as well as 6 newly created dictionaries based on systematic reviews of gender bias in recommendation letters, to compare recommendation letters generated for the 200 most historically popular "male" and "female" names in the United States. Study 1 used 3 different letter-writing prompts intended to accentuate professional accomplishments associated with male stereotypes, female stereotypes, or neither. Study 2 examined whether lengthening each of the 3 prompts while holding the between-prompt word count constant modified the extent of bias. Study 3 examined the variability within letters generated for the same name and prompts. We hypothesized that when prompted with gender-stereotyped professional accomplishments, ChatGPT would evidence gender-based language differences replicating those found in systematic reviews of human-written recommendation letters (eg, more affiliative, social, and communal language for female names; more agentic and skill-based language for male names). RESULTS: Significant differences in language between letters generated for female versus male names were observed across all prompts, including the prompt hypothesized to be neutral, and across nearly all language categories tested. Historically female names received significantly more social referents (5/6, 83% of prompts), communal or doubt-raising language (4/6, 67% of prompts), personal pronouns (4/6, 67% of prompts), and clout language (5/6, 83% of prompts). Contradicting the study hypotheses, some gender differences (eg, achievement language and agentic language) were significant in both the hypothesized and nonhypothesized directions, depending on the prompt. Heteroscedasticity between male and female names was observed in multiple linguistic categories, with greater variance for historically female names than for historically male names. CONCLUSIONS: ChatGPT reproduces many gender-based language biases that have been reliably identified in investigations of human-written reference letters, although these differences vary across prompts and language categories. Caution should be taken when using ChatGPT for tasks that have social consequences, such as reference letter writing. The methods developed in this study may be useful for ongoing bias testing among progressive generations of chatbots across a range of real-world scenarios. TRIAL REGISTRATION: OSF Registries osf.io/ztv96; https://osf.io/ztv96.

Loosening the GRIP (Gender Roles Inhibiting Prosociality) to Promote Gender Equality.

Prosociality is an ideal context to begin shifting traditional gender role stereotypes and promoting equality. Men and women both help others frequently, but assistance often follows traditional gender role expectations, which further reinforces restrictive gender stereotypes in other domains. We propose an integrative process model of gender roles inhibiting prosociality (GRIP) to explain why and how this occurs. We argue that prosociality provides a unique entry point for change because it is (a) immediately rewarding (which cultivates positive attitude formation), (b) less likely to threaten the gender status hierarchy, and therefore less susceptible to social backlash (which translates into less restrictive social norms), and (c) a skill that can be learned (which leads to stronger beliefs in one's own ability to help). Using the GRIP model, we derive a series of hypothesized interventions to interrupt the self-reinforcing cycle of gender role stereotyping and facilitate progress toward broader gender equality.

Did Descriptive and Prescriptive Norms About Gender Equality at Home Change During the COVID-19 Pandemic? A Cross-National Investigation.

Using data from 15 countries, this article investigates whether descriptive and prescriptive gender norms concerning housework and child care (domestic work) changed after the onset of the COVID-19 pandemic. Results of a total of 8,343 participants (M = 19.95, SD = 1.68) from two comparable student samples suggest that descriptive norms about unpaid domestic work have been affected by the pandemic, with individuals seeing mothers' relative to fathers' share of housework and child care as even larger. Moderation analyses revealed that the effect of the pandemic on descriptive norms about child care decreased with countries' increasing levels of gender equality; countries with stronger gender inequality showed a larger difference between pre- and post-pandemic. This study documents a shift in descriptive norms and discusses implications for gender equality-emphasizing the importance of addressing the additional challenges that mothers face during health-related crises.

Gender differences in grief narrative construction: a myth or reality?

Background: Narratives play a central role in the recovery process following death, and linguistic properties of grief narratives can serve as indicators of adjustment to loss. The present study examined whether bereaved men and women differ in how they discuss their loss, and how linguistic markers relate to psychological functioning. Positive associations were hypothesized between first-person singular pronoun use and psychological distress. Gender differences were expected for different emotion and social process words, and overall word use. Exploratory analyses were conducted to assess the relationship between linguistic markers and psychosocial outcomes for men and women separately. Method: 50 bereaved widow(er)s and parents (29 women, 21 men; M Age = 71.16 years, SD = 9.95) completed psychosocial self-report questionnaires and individual in-depth interviews. Grief narratives were analysed using Linguistic Inquiry and Word Count (LIWC), a software program that quantifies words into linguistic and psychological categories. Results: Contrary to our hypothesis, first-person pronoun use was not related to psychological distress. Although gender differences emerged in self-reported psychosocial outcomes, we failed to find the predicted gender differences in linguistic markers (emotion and social process words, overall word count). Exploratory analyses revealed additional associations between linguistic markers and psychosocial outcomes, and gender differences in these relationships. Notably, first-person pronoun use was related to heightened grief avoidance. Furthermore, various linguistic markers were associated with increased depression levels in females, but not males. In contrast, nonfluencies were positively associated with indicators of psychological distress in men only. Conclusion: In line with the gender similarities hypothesis, analyses suggest similarities between men and women's discussion of their grief experience. Associations between linguistic markers and psychological adjustment indicate that grief narratives contain meaningful indices of underlying health.

Antecedentes: Las narrativas juegan un rol central en el proceso de recuperación posterior al fallecimiento, y las propiedades lingüísticas de las narrativas del duelo pueden servir como indicadores de adaptación a la pérdida. El presente estudio examinó si es que hombres y mujeres en proceso de duelo difieren en como discuten su pérdida, y como los marcadores lingüísticos se relacionan con el funcionamiento psicológico. Se hipotetizó que existirían asociaciones positivas entre uso del pronombre singular en primera persona y distrés psicológico. Se esperaban diferencias de género para distintas palabras sobre emociones y palabras de proceso social, y uso de palabras en general. Se realizaron análisis exploratorios para evaluar la relación entre marcadores lingüísticos y resultados psicosociales para hombres y mujeres por separado.Método: 50 personas en proceso de duelo, tanto viudos como padres (29 mujeres, 21 hombres, M edad =71.16, Desviación Estándar = 9.95) completaron cuestionarios psicosociales de auto-reporte y entrevistas individuales en profundidad. Las narrativas del duelo fueron analizadas por Investigación Lingüística y Conteo de Palabras (LIWC por sus siglas en inglés) un programa de software que cuantifica palabras en categorías lingüísticas y psicológicas.Resultados: Al contrario de nuestra hipótesis, el uso del pronombre en primera persona no se relacionó con distrés psicológico. Aunque emergieron diferencias de género en los resultados psicosociales auto-reportados, no logramos encontrar las diferencias de género que fueron previstas en los marcadores lingüísticos (palabras sobre emociones y proceso social, conteo de palabras en general). Los análisis exploratorios revelaron asociaciones adicionales entre marcadores lingüísticos y resultados psicosociales, y diferencias de género entre estas relaciones. Cabe destacar que el uso de pronombres en primera persona estuvo relacionado con aumentada evitación del dolour por la pérdida. Además, varios marcadores lingüísticos se asociaron con aumento de niveles de depresión en mujeres, pero no en hombres. En contraste, las no fluencias tuvieron asociación positiva con indicadores de distrés psicológico solamente en los hombres.Conclusión: En línea con las hipótesis de las similitudes entre géneros, los análisis sugieren similitud de la discusión de la experiencia de duelo entre hombres y mujeres. Las asociaciones entre marcadores lingüísticos y adaptación psicológica indican que las narrativas sobre el duelo contienen índices significativos de salud subyacente.