Course Description
This course will introduce students to ways of thinking about how recent developments in AI systems powered by large language models (LLMs) shape everyday life and how to design such systems in a manner that can respect human values.
This course is a lecture-based course including interactive discussions and a final project. For the final project, students will form interdisciplinary groups of 2-3 members and will create an innovative Human-AI interaction system powered by LLMs (students can also use other deep learning architectures and modalities of their choice beyond text, including vision and audio, but will need to confirm with the instructor in advance). The system can be chosen from a variety of those discussed throughout the course. In the past, student final projects have included counterspeech generators powered by LLMs, an interactive sign-language learning system, and an image generating tool for food and menu design using prompts.
This is a highly interactive class: You’ll be expected to actively participate in activities, projects, assignments, and discussions.
This course will introduce students to ways of thinking about how recent developments in AI systems powered by large language models will shape everyday life and how to design such systems in manner that can respect human values. Students will read and discuss papers in Human-AI interaction powered by language models, including but not limited to:
- (1) Human-AI interactive systems powered by LLMs that work / clash with the strengths and weaknesses of human cognition,
- (2) Designing interactive, human-in-the-loop approaches in such systems, and
- (3) Supporting interpretability, transparency, trust, and fairness in AI tools supported by LLMs.
These topics will be explored in the context of real-world applications (e.g., “For Some Autistic People, ChatGPT Is a Lifeline”), through which students will learn how to think both optimistically and critically of what LLM-powered AI systems can do, and how they can and should be integrated into society.
Prequisites
At the minimum, students should have an intermediate proficiency in python programming. A basic knowledge of deep / machine learning, statistics, and prior coursework in Human-Computer Interaction (HCI) are a plus, but not required. This course will include in-class LLM tutorial sessions designed to help you with your course project.
Grading
- Class participation/attendance – 15%
- Reading responses (5 per semester) – 15%
- In-class reading presentations – 20%
- Final Project – 50%
- Team Formation – 0%
- Project Pitch Presentation – 10%
- Prototype + Midterm Presentation – 10%
- Final Project Presentation – 15%
- Final Prototype and Video – 15%
Late Policy
-
Reading Responses: No late days. All reading responses are due at 11:59 pm the day before class. Responses are meant to stir class discussions, so late submissions are not accepted.
-
In-class paper discussion questions: No late days. Questions are due at 4pm before the day of presentation.
-
In-class paper presentation: No late days. Slides are due at 4pm before the day of your presentation.
-
Prototype (midterm): No late days.
-
Final project presentation: No late days.
-
Final project: No late days.
Reading List
Week |
Date |
Topic |
Reading |
1
|
21-Aug
|
Introduction and Course Overview
|
Licklider, Joseph CR. “Man-computer symbiosis.” IRE transactions on human factors in electronics 1 (1960): 4-11. (read in class) |
Shyam Sankar. The Rise of Human Computer Cooperation. TED Talk Video, 2012 (12 mins). |
23-Aug
|
Primer on AI
|
Lubars, Brian, and Chenhao Tan. “Ask not what AI can do, but what AI should do: Towards a framework of task delegability.” In Advances in Neural Information Processing Systems, pp. 57-67. 2019. |
Xu, Anbang, Zhe Liu, Yufan Guo, Vibha Sinha, and Rama Akkiraju. “A new chatbot for customer service on social media.” In Proceedings of the 2017 CHI Conference on Human Factors in Computing Systems, pp. 3506-3510. 2017. |
Nityesh Agarwal. “Getting started with reading Deep Learning Research papers: The Why and the How”, a blog post at Towards Data Science (2018). |
2
|
28-Aug
|
LLM Overview
|
Shanahan, M. (2022). Talking about large language models. arXiv preprint arXiv:2212.03551. |
Zhao, W. X., Zhou, K., Li, J., Tang, T., Wang, X., Hou, Y., … & Wen, J. R. (2023). A survey of large language models. arXiv preprint arXiv:2303.18223. |
Can Computers Learn Common Sense?, The New Yorker, 2022 |
A mental health tech company ran an AI experiment on real users. Nothing’s stopping apps from conducting more. |
30-Aug
|
Primer on HCI
|
Amershi, Saleema, et al. “Guidelines for human-AI interaction.” Proceedings of the 2019 chi conference on human factors in computing systems. 2019. |
Yang et al., Re-examining Whether, Why, and How Human-AI Interaction Is Uniquely Difficult to Design (CHI 2020) |
Shneiderman, B., “Human-Centered Artificial Intelligence: Reliable, Safe & Trustworthy.” International Journal of Human-Computer Interaction 36, 6, 495-504. 2020. |
4
|
11-Sep
|
Prompting - 1
|
Zamfirescu-Pereira, J. D., Wong, R. Y., Hartmann, B., & Yang, Q. (2023, April). Why Johnny can’t prompt: how non-AI experts try (and fail) to design LLM prompts. In Proceedings of the 2023 CHI Conference on Human Factors in Computing Systems (pp. 1-21). |
AI Chains: Transparent and Controllable Human-AI Interaction by Chaining Large Language Model Prompts, Tongshuang Wu, Michael Terry, Carrie J Cai - CHI 2022 |
Skim: Liu, P., Yuan, W., Fu, J., Jiang, Z., Hayashi, H., & Neubig, G. (2023). Pre-train, prompt, and predict: A systematic survey of prompting methods in natural language processing. ACM Computing Surveys, 55(9), 1-35. |
13-Sep
|
Prompting - 2
|
PromptChainer: Chaining Large Language Model Prompts through Visual Programming, Tongshuang Wu, Ellen Jiang, Aaron Donsbach, Jeff Gray, Alejandra Molina, Michael Terry, Carrie J Cai - CHI 2022 |
Wei, Jason, et al. “Chain-of-thought prompting elicits reasoning in large language models.” Advances in Neural Information Processing Systems 35 (2022): 24824-24837. |
5
|
18-Sep |
LLM Tutorial - 2 (come to class with laptop) |
20-Sep
|
Fairness, Accountability, Transparency & Ethics in LLMs - 1
|
Prabhakaran, Vinodkumar, Ben Hutchinson, and Margaret Mitchell. “Perturbation sensitivity analysis to detect unintended model biases.” arXiv preprint arXiv:1910.04210 (2019). |
Goyal, Nitesh, et al. “Is Your Toxicity My Toxicity? Exploring the Impact of Rater Identity on Toxicity Annotation.” Proceedings of the ACM on Human-Computer Interaction 6.CSCW2 (2022): 1-28. |
Clark, Elizabeth, et al. “All that’s’ human’is not gold: Evaluating human evaluation of generated text.” arXiv preprint arXiv:2107.00061 (2021). |
6
|
25-Sep
|
Fairness, Accountability, Transparency & Ethics in LLMs - 2
|
Jakesch, M., Bhat, A., Buschek, D., Zalmanson, L., & Naaman, M. (2023, April). Co-writing with opinionated language models affects users’ views. In Proceedings of the 2023 CHI Conference on Human Factors in Computing Systems (pp. 1-15). |
Wenzel, K., Devireddy, N., Davison, C., & Kaufman, G. (2023, April). Can Voice Assistants Be Microaggressors? Cross-Race Psychological Responses to Failures of Automatic Speech Recognition. In Proceedings of the 2023 CHI Conference on Human Factors in Computing Systems. |
27-Sep
|
Fairness, Accountability, Transparency & Ethics in LLMs - 3
|
“Because AI is 100% right and safe”: User Vulnerabilities and Sources of AI Authority in India, Shivani Kapania, Oliver Siy, Gabe Clapper, Azhagu SP, Nithya Sambasivan - CHI 2022 |
Mendelsohn, J., Bras, R. L., Choi, Y., & Sap, M. (2023). From Dogwhistles to Bullhorns: Unveiling Coded Rhetoric with Language Models. arXiv preprint arXiv:2305.17174. |
11
|
11-Oct
|
LLM-Supported Health Care
|
Jo, E., Epstein, D. A., Jung, H., & Kim, Y. H. (2023, April). Understanding the benefits and challenges of deploying conversational AI leveraging large language models for public health intervention. (CHI 2023) |
Chen, S., Wu, M., Zhu, K. Q., Lan, K., Zhang, Z., & Cui, L. (2023). LLM-empowered Chatbots for Psychiatrist and Patient Simulation: Application and Evaluation. arXiv preprint arXiv:2305.13614. |
12
|
25-Oct
|
LLM Accessibility and Neurodiversity
|
Valencia, S., Cave, R., Kallarackal, K., Seaver, K., Terry, M., & Kane, S. K. (2023, April). “The less I type, the better”: How AI Language Models can Enhance or Impede Communication for AAC Users. In Proceedings of the 2023 CHI Conference on Human Factors in Computing Systems (pp. 1-14). |
For Some Autistic People, ChatGPT Is a Lifeline, WIRED, 2023 |
11
|
30-Oct
|
LLM-Supported Work: Writing
|
CoAuthor: Designing a Human-AI Collaborative Writing Dataset for Exploring Language Model Capabilities Mina Lee, Percy Liang, Qian Yang CHI 2022 |
Tale Brush: Sketching Stories with Generative Pretrained Language Models, John Joon Young Chung, Wooseok Kim, Kang Min Yoo, Hwaran Lee, Eytan Adar, Minsuk Chang CHI2022 |
01-Nov
|
LLM-Supported Work: Research
|
Hämäläinen, P., Tavast, M., & Kunnari, A. (2023, April). Evaluating large language models in generating synthetic hci research data: a case study. In Proceedings of the 2023 CHI Conference on Human Factors in Computing Systems (pp. 1-19). |
Park, Joon Sung, et al. “Social Simulacra: Creating Populated Prototypes for Social Computing Systems.” Proceedings of the 35th Annual ACM Symposium on User Interface Software and Technology. 2022. |
13
|
15-Nov
|
Text to Visual/ Audio
|
Lyu, C., Wu, M., Wang, L., Huang, X., Liu, B., Du, Z., … & Tu, Z. (2023). Macaw-LLM: Multi-Modal Language Modeling with Image, Audio, Video, and Text Integration. arXiv preprint arXiv:2306.09093. |
Make-An-Audio: Text-To-Audio Generation with Prompt-Enhanced Diffusion Models |
14 |
Nov 19-23 |
Thanksgiving Break |
15
|
27-Nov
|
Creative Applications
|
FaceChat: An Emotion-Aware Face-to-face Dialogue Framework |
Park, J. S., O’Brien, J. C., Cai, C. J., Morris, M. R., Liang, P., & Bernstein, M. S. (2023). Generative agents: Interactive simulacra of human behavior. |
29-Nov |
No Class, Work on Final Project |
16
|
04-Dec |
Final Presentation |
06-Dec |
Final Presentation |
Honor Code
The Virginia Tech Academic Honor System applies to all work. Be especially careful to avoid plagiarism, which includes using materials (ideas, code, designs, text, etc.) that you did not create without proper attribution. Students are encouraged to collaborate on project designs and evaluations, but the final exam is strictly individual. Any violations of the honor code will be reported.
Special Needs
If you have special needs or require special arrangements, you can reach out me privately.