Explanation is a topic of longstanding interest in philosophy and psychology, and has recently attracted renewed attention due to novel challenges in interpreting and interacting with relatively opaque AI systems. In this graduate seminar, we will study the science and engineering of explanations, combining perspectives from philosophy, psychology, AI, and the legal sciences. We will ask questions like: When do we ask for explanations? What makes a good explanation? How can we build machines that can understand and explain? This interdisciplinary seminar is co-taught by Thomas Icard (Philosophy) and Tobias Gerstenberg (Psychology). We will meet twice a week (Tuesdays and Thursdays 10:30am-11:50am) to discuss research articles from a range of disciplines. Students are expected to write responses based on their readings, lead the discussion on one of the papers, and actively participate in the discussion otherwise. As a final project, students will outline a novel study on explanation that makes an empirical, modeling, or theoretical contribution.

If you’re interested in taking this class, please fill out the course application form here by Wednesday, September 9. We will respond to all applicants by Friday, September 11.

---

Instructor info

Tobi Gerstenberg, Assistant Professor of Psychology

Office hours: Tuesday 1-2pm

Email: gerstenberg@stanford.edu

Thomas Icard, Assistant Professor of Philosophy and Computer Science (by courtesy)

Office hours: Wednesday 2-3pm

Email: icard@stanford.edu

Schedule

The class meets on Tuesdays and Thursdays from 10:30am to 11:50am. The readings will be made available through Canvas.

Week 1: Background

• 9/15: Woodward, J. (Winter 2019 Edition) Scientific Explanation, The Stanford Encyclopedia of Philosophy, Edward N. Zalta (ed.).
• 9/17: Lombrozo, T. & Vasilyeva, N. (2017). Causal explanation. Oxford Handbook of Causal Reasoning, 415-432.

Week 2: Explanation and understanding in science

• 9/22: Trout, J. D. (2007). The Psychology of Scientific Explanation. Philosophy Compass, 2(3), 564-591.
  • optional: Craver, C. (2014). The ontic account of scientific explanation.
• 9/24: Grimm, S. R. (2010). The goal of explanation. Studies in History and Philosophy of Science Part A, 41(4), 337-344.
  • optional: de Regt, H. & Dieks, D. (2005). A contextual approach to scientific understanding, Synthese volume 144, 137–170.

Week 3: Individual-level function of explanation

• 9/29: Lombrozo, T. (2006). The structure and function of explanations. Trends in Cognitive Sciences, 10(10), 464-470.
  • optional: Keil, F. C. (2006). Explanation and Understanding. Annual Review of Psychology, 57(1), 227–254.
• 10/1: Liquin, E. G. & Lombrozo, T. (2020). A functional approach to explanation-seeking curiosity. Cognitive Psychology, 119, 101276.

Week 4: Communication

• 10/6: Hilton, D. J. (1990). Conversational processes and causal explanation. Psychological Bulletin, 107(1), 65-81.
  • optional: Turnbull, W. (1986). Everyday explanation: The pragmatics of puzzle resolution. Journal for the Theory of Social Behaviour, 16(2), 141-160.
• 10/8: Kirfel, L., Icard, T. F., & Gerstenberg, T. (2020). Inference from explanation. PsyArXiv.
  • optional: Potochnik, A. (2016). Scientific explanation: Putting communication first. Philosophy of Science, 83(5), 721-732.

Week 5: Formal theories of explanation

• 10/13: Halpern, J. Y. & Pearl, J. (2005). Causes and explanations: A structural-model approach. Part II: Explanations. The British Journal for the Philosophy of Science, 56(4), 889-911.

• 10/15: Bareinboim, E., Correa, J., Ibeling, D., & Icard, T. (2020). On Pearl’s hierarchy and the foundations of causal inference. Probabilistic and Causal Inference: The Works of Judea Pearl, ACM. (Please read sections 1.1-1.3. Sections 1.4-1.5 are optional.)

Week 6: NLP & Vision

• 10/20: Vig, J., Gehrmann, S., Belinkov, Y., Qian, S., Nevo, D., Singer, Y., & Shieber, S. (2020). Causal mediation analysis for interpreting neural nlp: The case of gender bias. arXiv preprint arXiv:2004.12265.
  • optional: Murty, S., Koh, P. W., & Liang, P. (2020). ExpBERT: Representation Engineering with Natural Language Explanations. arXiv preprint arXiv:2005.01932.
  • optional: Hancock, B., Bringmann, M., Varma, P., Liang, P., Wang, S., & Ré, C. (2018). Training classifiers with natural language explanations. In Proceedings of the conference. Association for Computational Linguistics Meeting (pp. 1884).
• 10/22: Hendricks, L. A., Akata, Z., Rohrbach, M., Donahue, J., Schiele, B., & Darrell, T. (2016). Generating visual explanations. In European Conference on Computer Vision (pp. 3-19).
  • optional: Hendricks, L. A., Hu, R., Darrell, T., & Akata, Z. (2018). Grounding visual explanations. In European Conference on Computer Vision (pp. 269-286).

Week 7: Reinforcement learning and action

• 10/27: Buesing, L., Weber, T., Zwols, Y., Racaniere, S., Guez, A., Lespiau, J.-B., & Heess, N. (2018). Woulda, coulda, shoulda: Counterfactually-guided policy search. arXiv preprint arXiv:1811.06272.
  • optional: Schulam, P. & Saria, S. (2017). Reliable decision support using counterfactual models. In Advances in Neural Information Processing Systems (pp. 1697-1708).
  • optional: Oberst, M. & Sontag, D. (2019). Counterfactual off-policy evaluation with Gumbel-Max structural causal models. arXiv preprint arXiv:1905.05824.
• 10/29: Karimi, A.-H., Barthe, G., Schölkopf, B., & Valera, I. (2020). A survey of algorithmic recourse: definitions, formulations, solutions, and prospects. arXiv preprint arXiv:2010.04050.
  • optional: Venkatasubramanian, S. & Alfano, M. (2020). The philosophical basis of algorithmic recourse. In Proceedings of the 2020 Conference on Fairness, Accountability, and Transparency (pp. 284–293).

Week 8: Prediction vs explanation

• 11/3: no class because of election day

• 11/5: Yarkoni, T. & Westfall, J. (2017). Choosing prediction over explanation in psychology: Lessons from machine learning. Perspectives on Psychological Science, 12(6), 1100–1122.

  • optional: Shmueli, G. (2010). To explain or to predict?. Statistical Science, 25(3), 289–310.

Week 9: Legal dimensions

• 11/10: Wachter, S., Mittelstadt, B., & Russell, C. (2017). Counterfactual explanations without opening the black box: Automated decisions and the GDPR. SSRN Electronic Journal.
  • optional: Mittelstadt, B., Russell, C., & Wachter, S. (2019). Explaining explanations in AI. In Proceedings of the conference on fairness, accountability, and transparency (pp. 279–288).
• 11/12: Doshi-Velez, F., Kortz, M., Budish, R., Bavitz, C., Gershman, S., O’Brien, D., Schieber, S., Waldo, J., Weinberger, D., & Wood, A. (2017). Accountability of AI under the law: The role of explanation. arXiv preprint arXiv:1711.01134.
  • optional: Narayanan, M., Chen, E., He, J., Kim, B., Gershman, S., & Doshi-Velez, F. (2018). How do humans understand explanations from machine learning systems? an evaluation of the human-interpretability of explanation. arXiv preprint arXiv:1802.00682.

Week 10: Project presentations

• 11/17
• 11/19

General information

What to expect?

In “A Vision for Stanford”, university president Marc Tessier-Lavigne states that Stanford wants to be

“an inspired, inclusive and collaborative community of diverse scholars, students and staff, where all are supported and empowered to thrive.”

Let’s try our best together in this seminar to make this happen!

What you can expect from us

We will …

• provide an introduction to the philosophy and psychology of explanation in Week 1.
• help facilitate the discussion in subsequent sessions.
• provide feedback on your final papers.
• be available to talk during office hours.
• send announcements via Canvas to guide the reading for each upcoming week.

What we expect from you

You will …

• attend all of the classes and participate in class discussion.
• lead the discussion for one class.
• write short reaction posts based on the readings and upload each paper on Canvas by 10pm (at the latest) the night before class.
• write a final paper.
• present your final paper in class in Week 10.
• students who audit the class are expected to write reaction posts and participate in discussion in class.

Grading

• 25% participation in class
• 25% leading discussion in one class
• 25% reaction posts to readings
• 25% final paper

Reaction posts

Here are some guiding thoughts on how to write a good reaction post:

• Express your opinion rather than summarize the paper(s).
• Try to connect the ideas expressed in the paper to concrete every-day experiences.
• Identify strengths and weaknesses of the paper.
• Relate different papers to each other.
• Ask questions that go beyond what the paper discusses (what’s missing, where should we go next)?

The reaction posts should be concise (~ two paragraphs per paper), and are due the night before class (submitted via Canvas).

Final paper

The final paper may be one of the following three:

1. An empirical project proposal.
2. A literature review based on one of the class topics.
3. A theoretical essay.

The final paper (1000–2000 words) will be due on November, 19th at 10pm.

Policies

Please familiarize yourself with Stanford’s honor code. We will adhere to it and follow through on its penalty guidelines.

Support

Students who may need an academic accommodation based on the impact of a disability must initiate the request with the Office of Accessible Education (OAE). Professional staff will evaluate the request with required documentation, recommend reasonable accommodations, and prepare an Accommodation Letter for faculty dated in the current quarter in which the request is being made. Students should contact the OAE as soon as possible since timely notice is needed to coordinate accommodations. The OAE is located at 563 Salvatierra Walk (phone: 723-1066, URL: http://oae.stanford.edu).

Stanford is committed to ensuring that all courses are financially accessible to its students. If you require assistance with the cost of course textbooks, supplies, materials and/or fees, you should contact the Diversity & First-Gen Office (D-Gen) at opportunityfund@stanford.edu to learn about the FLIbrary and other resources they have available for support.

Stanford offers several tutoring and coaching services:

• Academic Skills Coaching
• Tutor for learning differences
• English language learners
• Hume Center for Writing and Speaking

Feedback

We welcome feedback regarding the course at any point. Please feel free to talk with us after class, come to office hours, email us, or leave anonymous feedback using this online form.