AI evidence extraction

Co-designing AI Education Curriculum with Cross-Disciplinary High School Teachers

Benjamin Xie, Parth Sarin, Jacob Wolf, Raycelle C. C. Garcia, Victoria Delaney, Isabel Sieh et al. · Proceedings of the AAAI Conference on Artificial Intelligence · 2024

AI-generated evidence extraction, verified across multiple analytical personas. Not a substitute for the peer-reviewed original.

6/10

Relevance

2/4

Quality (LMQS)

Evidence

Citations

10.75

FWCI

Top 10%

Impact

This is an AI analysis. Read the peer-reviewed original at the publisher: https://doi.org/10.1609/aaai.v38i21.30360

Methodology & findings

Study design

Qualitative co-design study with thematic analysis.

Sample

N = 8, 3 groups

Primary method

Thematic analysis using Braun and Clarke (2006) approach. Two researchers independently open-coded recordings and transcripts, developed affinity maps, and created a codebook that was then applied to remaining data. Sensitizing concepts used: teachers' disciplines and backgrounds, teachers' considerations of student perspectives, and teachers' prior AI experiences. No quantitative statistical analysis was performed.

Main result

Teachers considered AI from both technical and ethical perspectives when teaching across disciplines. The study found that "teachers considered AI from both technical and ethical perspectives, with the importance of each perspective varying by discipline. Two humanities teachers identified ways for 'dual exploration' where students could apply discipline-specific analytic frameworks to AI generated art to think more critically about the frameworks as well as about AI." Additionally, "teachers considered AI tools as complementary to students' skills as well as capable of enabling students to do something they otherwise could not. Teachers framed limitations of AI tools as discussion or reflection opportunities, as well as potential distractions."

Reports effect sizes.

Research paradigm

Qualitative/Interpretivist

Author conclusions

The authors conclude that "integrating these perspectives in ways that are compatible to different disciplines" is necessary, and that "we must prepare teachers to teach about AI from technical and ethical perspectives." They also emphasize that "there are opportunities to design for 'dual exploration' of disciplinary learning alongside learning about AI and its limitations," and suggest that "by partnering with teachers throughout the design process, we can realize learning experiences where teachers have relevant preparation to teach about AI using AI tools to support disciplinary learning and/or student reflection, while also ensuring students' safety and agency given the unpredictable nature and biases of AI tools."

Risk of bias

Selection bias: Teachers recruited had prior relationships with research team members; Social desirability bias: Participants may have provided responses researchers wanted to hear due to existing relationships; Attrition: 13 teachers began study, 4 dropped due to time constraints, 1 did not return consent form; Geographic/cultural bias: Study conducted in Western, educated, industrialized, rich, democratic (W.E.I.R.D.) context; Researcher positionality: Researchers had varying degrees of prior involvement with teacher participants; Social Desirability Bias: Prior relationships between researchers and teachers (4 teachers were prior study participants, 1 had taken researcher's class, 2 from school where researcher volunteered); Selection bias: Teachers recruited were already interested in AI education; Attrition: 5 of 13 initially recruited teachers did not complete study (4 due to time constraints, 1 did not return consent form); WEIRD context bias: Study conducted in Western, educated, industrialized, rich, democratic context, limiting generalizability; Potential for confirmation bias in qualitative coding despite independent coding by two researchers; Social Desirability Bias due to prior relationships between researchers and participants; Selection bias - recruited teachers with prior connections to researchers; Researcher bias - two coders had prior relationships with some participants; WEIRD (Western, Educated, Industrialized, Rich, Democratic) context bias; Attrition - 5 of 13 originally recruited teachers did not complete the study

Limitations

The authors acknowledge that "we recruited teachers that authors had some prior relationship with
Four teachers had previously been participants in other studies with a researcher, one teacher had taken a class taught by that same researcher, and two teachers were from a high school that another researcher volunteered at
This could have led to Social Desirability Bias." Additionally, they note "this research was conducted in a Western, educated, industrialized, rich and democratic (W.E.I.R.D.) societal context
While this is common amongst neighboring research communities, most of the world is not W.E.I.R.D
Most teachers in our study did work at Title I schools serving Black, Hispanic, and Latinx students..
Nevertheless, we acknowledge this bias to create space for future work in less W.E.I.R.D

Open questions raised

Future work to integrate technical and ethical perspectives of AI in ways compatible with different disciplines
Need to prepare teachers to teach about AI from both technical and ethical perspectives, particularly for non-CS teachers
Opportunities to design for 'dual exploration' of disciplinary learning alongside learning about AI limitations
Research needed in less W.E.I.R.D. (non-Western) contexts
Investigation of how teachers can be supported to teach about AI with confidence across disciplines
Need to make technical and ethical perspectives of AI more compatible with different disciplines

Data: Not reported as publicly available.Code: Not mentioned.Extracted from: pdfAgreement 55%

Explore related topics

AI Literacy Training Doctoral Training Design Science