This course is structured around seven Big Ideas for Human AI Interaction. We will focus on each Big Idea for about two weeks of class time, but these ideas are woven into the entire fabric of the course. The assignments and projects in this class are also centered around these Big Ideas.
First, we will look at the transformation of AI over the past few years based on the development of large foundational models and what it means for human AI interaction. In this course, we will focus on large language models, but in this introductory segment, we will also talk about other large, foundational models such as those transforming text into images.
2. Grappling with the interaction problem – journey mapping, mapping rewards and risks (RICE), interaction paradigms (agents, tools, mixed initiative)
Designing meaningful human interaction, requires and understanding both technology and the human needs that the AI is going to solve. In this segment, we will discuss a few systematic ways of understanding human needs and translating them into machine capabilities. We will also talk about different interaction paradigms such as: proactive agents, reactive tools etc.
Large language, models are quickly becoming the de facto way to design interesting human AI interaction applications. We will dive deep into using large language models and prompting them for good results.
Today, it is easy to create an interactive AI system that mostly works, that is just broken enough to be useless in real life. We will discuss how to design interactive AI to gather data, and to improve the usefulness of the system over time. We will also look at how data practices can be extractive and otherwise harmful of the people data is gathered from and discuss some mitigating strategies.
Human – AI teaming is an emergent interaction paradigm for an AI applications. How do you design for human AI teaming that is better than the sum of their parts?
The role of human computer interaction, researchers and designers is rapidly shifting with the growth of AI. As it becomes easier to implement interactive systems, the focus is shifting elsewhere. In this module, we will look at three emerging roles: the Community Partner, the HAI Engineer, and the AI-UX Researcher, and how you can prepare yourself for these roles.
Great human AI interaction is a synthesis of multiple principles. In this final module, we will talk about how to integrate all you have learned while making trade-offs using both data and your expert judgment. Throughout the course, we will try and invite leading researchers, designers, and thinkers of human AI interaction to deliver (usually remote) guest lectures. Attendance in these lectures is mandatory and counts towards your course grade.
You are allowed to use ChatGPT/other generative AI tools to:
- Create images
- Write/debug code
- Brainstorm ideas
- Better understand a topic (at your own risk – beware hallucinations!) … except where it is specifically disallowed.
All generative AI use MUST be disclosed in work that you turn in. Not disclosing use of AI is an honor code violation.
You may not use any generative AI tool, website, or aid in …
- completing quizzes in this class. The only resources allowed are the course slides, and notes you took yourself.