Authors:
(1) Evan Shieh, Young Data Scientists League ([email protected]);
(2) Faye-Marie Vassel, Stanford University;
(3) Cassidy Sugimoto, School of Public Policy, Georgia Institute of Technology;
(4) Thema Monroe-White, Schar School of Policy and Government & Department of Computer Science, George Mason University ([email protected]).
Table of Links
1.1 Related Work and Contributions
2.1 Textual Identity Proxies and Socio-Psychological Harms
2.2 Modeling Gender, Sexual Orientation, and Race
3 Analysis
4 Discussion, Acknowledgements, and References
SUPPLEMENTAL MATERIALS
A OPERATIONALIZING POWER AND INTERSECTIONALITY
B EXTENDED TECHNICAL DETAILS
B.1 Modeling Gender and Sexual Orientation
B.3 Automated Data Mining of Textual Cues
B.6 Median Racialized Subordination Ratio
B.7 Extended Cues for Stereotype Analysis
C ADDITIONAL EXAMPLES
C.1 Most Common Names Generated by LM per Race
C.2 Additional Selected Examples of Full Synthetic Texts
D DATASHEET AND PUBLIC USE DISCLOSURES
D.1 Datasheet for Laissez-Faire Prompts Dataset
B.5 Subordination Ratio
For power-laden prompts, we define the subordination ratio as the proportion of a demographic observed in the subordinate role compared to the dominant role:
This allows us to focus on relative differences in the portrayal of characters when power-laden prompts are introduced. If the subordination ratio is less than 1, we observe dominance; if the subordination ratio is greater than 1, we observe subordination; and if the subordination ratio is 1, then the demographic is neutral (independent from power dynamics).
B.6 Median Racialized Subordination Ratio
Characters generated by the language models bear names with low racial likelihood for all races except White (as shown in Fig. 3a). Therefore, overall subordination will be predominantly influenced by the large volume of Anglicized names. We are more interested in examining how the subordination ratio changes as the threshold varies. If no subordination effect exists, the null hypothesis states that thresholding would not have an impact and the ratio should concentrate around 1 given sufficient sample size.
However, setting a “one-size-fits-all” threshold is inherently subjective, and would fail to take into account real differences that affect name distributions between racial groups, including historical differences in periods of migration and assimilation (voluntary, involuntary, or restricted) [33, 37].
To address this, we introduce the median racialized subordination ratio to quantify subordination across a range of possible racial thresholds. First, we control for possible confounding effects of textual cues beyond name by conditioning on gender references (pronouns, titles, etc.). Then, for each intersection of race and gender we take the median of all subordination ratios for names above a variable likelihood threshold t as defined below:
With sufficiently granular t, this statistic measures subordination while taking the spectrum of racial likelihoods into account. For our experiments, we set t ∈ [1, 2, … 100]. Using the median controls for possible extremes; however, we nevertheless observe astonishingly high rates of subordination (see Fig. 3c) even though this approach conservatively underestimates perceived subordination (or domination) for racial groups with a high proportion of Anglicized names (e.g., Black names [55]).
This paper is