Huge huge kudos to @anthropic.com
for releasing this detailed overview of how students are using AI.

Some pretty concerning trends here, such as students completely off-loading higher-level reasoning to LLM and mostly prompting for direct answers.

www.anthropic.com/news/anthrop...

12/12 Paper written by me, @olalandersson.bsky.social , @perklasengstrom.bsky.social and Stefan Eriksson.

11/12 To conclude, our study highlights the “pioneer penalty” for minority faculty in a system with strong student biases and a homogenous student body. For external validity it remains critical to think about who are the "role model compliers" in the longer run, inside and outside the university.

10/ Herein lies an admission tradeoff. Absent admission rules, minority teachers attract fewer students. But equalizing class sizes via common merit-based admission concentrates weaker students on minority faculty. Both these outcomes may negatively impact underrepresented teachers' careers.

9/12 While the gender bias is easily nudged away, same-ethnicity preferences are more robust. When set as the default, underrepresented-gender instructors get balanced or near-balanced selection. Ethnic biases, however, are much stronger and persist despite default assignment.

8/12 Biases are intersectional. Female students strongly favor native female instructors but avoid immigrant male teachers entirely. Male students mirror this, although less strongly, preferring native male instructors over immigrant women.

7/12 We find that students show clear biases in class selection. Women prefer female teachers, men prefer male teachers, and there’s a strong bias against immigrant teachers. These biases are especially strong among younger, lower-performing, and female students.

6.5/12 Methodological point: The students did not know that the teacher names were fictive, and the setup was similar to how students regularly design their elective class schedules (it was a "natural field experiment" to use Harrison and @johnlist.bsky.social terminology).

6/12 To explore, we conducted a field experiment at a prestigious Swedish university. Students already admitted to the program (40% women, high SES) ranked elective seminar classes based on fictive teacher names signaling gender (male/female) and ethnicity (native/MENA background).

5/12 We hypothesized that advertising a diverse lineup of teachers may lead to internal student sorting that unintentionally disadvantages minority teachers. The mechanism is that the more socially represented teachers cream-skim the best students, depending on the admission system.

4/12 A key conceptual point is that many university courses are elective within a set program. Role models do not primarily attract students from outside the university but within a pool of homogenous students. Thus, if teachers “compete” using role model mechanisms, the board is already skewed!

3/12 An optimistic take on the role-model literature is that underrepresented teachers will primarily attract skilled underrepresented students, who will replace less-skilled majority students. However, the outcome depends on which students demand role models and how admission is structured.

2/12 Women and minorities remain underrepresented in fields like economics and STEM. Many departments aim to address this gap. As affirmative action in admissions is prohibited, an alternative is advertising the presence of minority faculty when students choose courses—a "role model"-strategy.

1/12 New working paper! The pioneer penalty in higher education seems to be real. Our new study examines how student biases influence course selection and provides cautionary lessons for role model advertising.

Paper: osf.io/preprints/so...

Trying out bluesky now that overleaf died on me. Hello world!