FS 2025 Human Computer Interaction in AI

• Module:  22MI0038
• Instructor: Dan M. Russell
• Schedule: Wednesday 10:15-12:00
• Room: BIN-1.D.29 

Course description

As you know, AI and ML are now essential parts of many systems that are currently being built-they're everywhere.  As such, what should HCI people know about the possibilities and potential drawbacks of building AI systems?

Understanding the human side of AI/ML based systems requires understanding both how the system-side AI works, but also how people think about, understand, and use AI tools and systems. What do they believe about AI?  How do they know this?  

This course will cover what AI components and systems currently exist, how to design and build usable systems with AI components, along with how the mental models of AI/ML tools operate. These models lead to user expectations of how AI systems function, and ultimately, to design guidelines that avoid disappointing end-users by accidentally creating unintelligible AI tools. We'll also cover the ethics of AI, including data collection, algorithmic and data fairness considerations, along with other risks of AI.
 

High-level topics covered in the course:    

* The human aspects of designing and building AI/ML systems - practice and theory
     - How people understand AI systems:  AI and mental models
     - examples/cases of AI system UX

*  Designing for AI failures and Feedback to Users
     - guardrails and failure modes
     - details on what has worked, what hasn’t worked, and why

* Data, Knowledge, Fairness, and Ethics
        - AI Ethics of Actions, Fairness, Social Acceptability, and Trust
        - analysis methods to understand HAI data with respect to fairness

*  Interpreting and Explaining AI Algorithms and Systems
      - Building AI/ML with humans in the loop / AI in the loop
      -  understanding spoken language; written language;
      - generating language; conversations;
      - large language models such as ChatGPT

* Computer perception: recognition, classification, uses (and misuses)

* AI & Art: Synthesis systems for creativity, music, imagery
      - prompt engineering (what it is and how to do it)
      - limits to AI & Art

* Where does the future of AI/ML and HCI lead?
      - HAI: how to design and build real systems
      - Or is AI/ML in an existential crisis?
      - If so, how and why?  What can we do about it?


The course will include several in-class exercises with existing AI tools. Materials for the exercises will be made available in the class via a public class website.  Participants should plan on bringing a laptop to the class (or be willing to partner with someone who has one) in order to do the in-class work. 

Interned Learning Outcomes

Students in the class will acquire knowledge about the human-centered design and core implementation knowledge needed to build operational computational systems that have AI subsystems as essential components. The learning outcomes will be (a) detailed knowledge of current AI systems and their expression in user interfaces; (b) knowledge about how to go through a human-centered design process to get to implementation; (c) the key factors of ethical and fairness issues in AI systems; (d) deep knowledge about how current AI systems came to be, and what to expect in their development over the next five years.