Using AI in Research | Stanford CS193T Guest Panel with Mark Algee-Hewitt, Megan Ma & Markus Covert

Using AI in Research | Stanford CS193T Guest Panel with Mark Algee-Hewitt, Megan Ma & Markus Covert

Title history
  • Using AI in Research | Stanford CS193T Guest Panel with Mark Algee-Hewitt, Megan Ma & Markus Covert — observed on 20261112-1909 UTC (YouTube)
🎙 Stanford Online 👥 1.3M 📅 November 12, 2026 ⏱ 71 min 👁 254 📄 expert opinion 🧭 2026-11-12
Available in: English (current) Français

Keywords

AI applicationsstyle transferlegal evaluationwhole-cell simulationeducation

Summary

The video is a recorded guest panel from Stanford’s CS193T course, hosted by Cynthia Bailey, featuring Professor Mark Algee-Hewitt (English/Digital Humanities), Megan Ma (Law/LIFT Lab), and Professor Markus Covert (Bioengineering). Mark Algee-Hewitt discusses his work on computational analysis of literature, including a project analyzing how AI style affects meaning across disciplines. He demonstrates a style transfer experiment where philosophy articles are rewritten as computer science, revealing a measurable shift in topic distribution. Megan Ma presents the LIFT Lab’s initiatives in legal AI, covering evaluation methods (preference rankings vs. rubrics), AI-based deposition simulators for training lawyers, and the use of AI in legal practice. Markus Covert shares his work on whole-cell simulation and deep learning for cell tracking, then raises broader questions about AI in education, human flourishing, and the need for scientists to have time for deep thinking. The panel then addresses audience questions on topics like AI for recommendation letters, transparency norms, and which skills students should preserve. The discussion emphasizes responsible AI use, the importance of human judgment, and the ethical considerations of integrating AI into academic and professional settings.

182 words

Critical Evaluation

Value of the Information & Strength of the Argument

The video provides significant value by offering concrete, field-specific examples of AI integration in research, moving beyond generic hype. Mark’s style transfer project is a novel demonstration of the subtle relationship between linguistic style and meaning, using computational methods to reveal how AI can alter even the topic of a text. Megan’s finding that preference rankings outperform rubrics in legal AI evaluation challenges conventional assumptions and has practical implications for building better AI tools. Her deposition simulator is an innovative application of AI for experiential learning in law. Markus raises compelling questions about the purpose of education and the risk of AI-induced over-reliance, citing the MIT report to caution against uncritical adoption. The arguments are presented logically and supported by the speakers’ authority and ongoing research, though they are more exploratory than conclusive. The panel format allows for diverse perspectives, but the lack of deep-dive technical details or rigorous statistical evidence limits the argumentation to illustrative rather than definitive.

Scientific Rigor, Source Quality, Title Accuracy

The video is produced by Stanford Online and features faculty from respected departments, lending inherent credibility. The speakers reference their own published research (e.g., JudgmentBench) and reports (the MIT AI education report) but do not provide direct citations or URLs in the video description. The description includes links to Stanford Online’s AI courses, the course archive, and the YouTube playlist, which are relevant but not direct references to the discussed studies. The title ‘Using AI in Research’ accurately encapsulates the broad topic, and the specifics of each speaker’s domain are clear. The discussion is transparent about the exploratory nature of some findings, and the speakers openly discuss limitations (e.g., Mark’s 43% style classification accuracy). Overall, the source quality is high, but the lack of detailed citations in the description slightly reduces the rigor. No audience comments were provided for analysis.

314 words

Title / Content Match

The title accurately reflects the content: a panel discussion on using AI in research across different disciplines, with specific examples and insights.

Quality & Reliability

8/10

The panel features three distinguished Stanford researchers with deep expertise in their fields, and the discussion is grounded in their ongoing research and real-world applications. However, it is a conversational panel rather than a peer-reviewed presentation, and some claims (e.g., the MIT report findings) are cited without full details. Overall, the information is reliable and credible.

Chapters

Cited Sources

Contribution & Novelties

This panel contributes a unique cross-disciplinary perspective on AI in research, illustrating how AI is not a one-size-fits-all tool but adapts to the epistemic norms and challenges of different fields. Mark Algee-Hewitt’s style transfer experiment is an original approach to interrogating the relationship between style and meaning, suggesting that AI-generated text can subtly shift semantic content—an insight with implications for authorship and authenticity. Megan Ma’s emphasis on preference ranking over rubrics offers a novel evaluation paradigm for ‘judgment-rich’ domains where ground truth is ambiguous. Markus Covert’s reflection on the MIT report and the ’time to think’ crisis frames AI adoption as a potential threat to deeper cognitive skills, raising critical questions for education and research culture. These ideas collectively highlight the need for human oversight and ethical safeguards in AI deployment.

Pour aller plus loin :

  • Digital humanities — Pertinent to Mark Algee-Hewitt’s work on computational literary analysis.
  • Topic model — The method used to visualize and analyze article content in the style project.
  • Preference learning — Relevant to Megan Ma’s JudgmentBench and preference-ranking approach in legal AI.
  • Whole-cell simulation — Key concept in Markus Covert’s research; further background can be found in bioengineering literature.

195 words

Radar Profile

The scores indicate a strong balance between quantity and quality of information (9/9), with slightly lower technical depth (7) due to the accessible panel format rather than in-depth mathematical detail. Reliability is high (8), reflecting the expert status of the speakers, while the overall rating of 4/5 aligns with the informative yet exploratory nature of the discussion.

Reliability 8/10