Human-Centered AI

Assignments:

Short Assignment #1: What will learning be like in an AI world?

For this assignment we’d like you to read/view/listen to some pieces on the topic of what it will mean to be a learner in a world where strong AI exists, and to write an essay on the topic. 

• Step 1: Pick at least one [of an instructor-provided list of written and media sources] to review in detail.
• Step 2: Find at least one other recent essay/video/podcast on this topic.  (Be sure to tell us what it is and provide a link to it.) 
• Reflect on what you’ve seen (both your chosen essay from Step 1 and your found essay from Step 2) and reply to the following prompts:
  • Write up your initial thoughts after your readings. What are some of the things that surprised you? What is something you disagree with? What questions arose for you? (This section will probably take up a paragraph to half a page.)
  • How do you think AI will change the way you go through your Stanford education? This section will probably take up 1 page, maybe multiple pages.
  • After what you’ve learned in the readings/videos, what are three steps you would take in order to make sure that the education you are getting is human-centered, ethical and fair? This section will probably take up a paragraph to half a page.
  • Most importantly, in a paragraph, tell us how you believe university level education will be changing as AI systems become a larger part of our lives and our learning experiences. 
  • We expect in total roughly 2 to 5 pages of your thoughts and considerations.
• Step 4: Reflect on your process of doing the whole assignment. How did you search for and find the piece from Step 2? Why did you choose that one? If you used an AI tool such as Grammarly, Gemini, Claude, or ChatGPT to help, describe the tools and prompts you used, and how you incorporated the output with your own writing. 

Short Assignment #2: Creating art using AI?

The new generation of text-prompt-guided image-generation models gives us the ability to rapidly explore a visual design space.  By allowing users to give the system simple text prompts and quickly create images through these descriptions, the prompts act as a reflective design material aiding exploration and collaboration. 

It’s pretty clear that artwork creators (e.g., designers) create prompts based on their implicit understanding of the underlying model–their mental model of what’s going on. What the system gives back as output can spark new ideas, allow for fast iteration of the prompts, and novel ways to generate new ideas.  

In this assignment we explore the question, “How does using prompt-based image generation change the design process of non-professional designers?”  In particular, does working with such a tool give you a sense of co-creation while working together? 

• PART 1:  You have been tasked to design a party invitation to your dorm’s spring party. You may create multiple versions of the invitation, and then choose a final design to be sent to all of your friends. Your invitation must contain: (a) a time, date, and location for the party (please make something up).  (b) an image that is recognizably somewhere on campus (a dorm, Memchu, White Plaza, really any place we’d recognize as clearly on campus).  (c) have some kind of border around the perimeter (like a picture frame). 
  • Task: Using the text-driven image synthesis system of your choice (Dall-E, Parti, image, Midjourney, etc.) AND Google Slides as your image editing system, create a design for your party invitation that has a spring theme. Think about who your target audience is, how you might catch their attention, what would convince them to come to a party and create your design accordingly. Be sure that you are making conscious design choices. Your invitation should be in a classic letter layout (8.5 X 11, or A4).  
• PART 2: Now, do the same task again except that we want you to design a card for a party to celebrate the Apollo moon landing.  Apollo 11 (July 16–24, 1969) was the American spaceflight that first landed humans on the Moon.
  • Task:  For this card, please use regular search engine image search (you may use Google, Bing, Tineye, Unsplash, anything that works for you). Do NOT use an image synthesis system–just regular image search. Again, think about who your target audience is, how you might catch their attention, what would convince them to come to a party and create your design accordingly. Be sure that you are making conscious design choices. Your invitation should be in a classic letter layout (8.5 X 11, or A4) that has text saying when/where the party will be held. Don’t use any photos of real people in your designs. Please don’t use copyrighted imagery (see Unsplash for material free of copyright).
• PART 3:  Write a 1 page reflection about your design process. Who did you design for? What are their needs, how can you catch their attention? How did you integrate this in your work? What did you learn about how the particular synthesis system prompts worked (or didn’t work)? Which design process did you prefer and why? How did your mental model of prompt-driven image synthesis change? Which was the simpler way to do the task? Be sure to tell us what surprised you, what you discovered, and what was disappointing in the process. What worked? What didn’t work?  What do you think? 

Short Assignment #3: Wizard of Oz Study

We often need to build AI systems that intrinsically have humans as moderators, guides, or drivers. In this assignment, you’ll prototype your idea for a human-in-the-loop system and run a short Wizard-of-Oz study.  

Goals: (1) To explore what it means to get a human-in-the-loop system to operate in a reasonable way. (2) To use a Wizard of Oz study method and get experience with it. 

Your assignment:  With a partner or two, set up a Wizard of Oz study for a fictitious system that needs to have a human-in-the-loop.  

• You have a number of options here to actually do the work. We encourage you to build something to support your WoZ study, but don’t spend too much time on it as we’re not judging your code, but the quality of the study itself. 
• A no-code solution would be to use an open phone connection–the participant could then talk to the AI back-end in a free conversation. Another way would be to use regular text-messaging to text back and forth between the participant and the simulated AI system.
• Ideally, the “simulated AI” person should not know the test participant (too much shared context). 
• In any case, you need to run a 15 minute study exploring different situations that will stress the system in ways that you (as the designer) might not have anticipated. 
• As you do the study, video record the experiment and then edit together a 1 minute video that summarizes what you did and what you found.

Create a video summary of your findings:  When you create your video, be sure to: (a) describe / show what your WoZ setup is; (b) show a typical interaction between the participant and the “system” is; (c) summarize your findings about the utility (or not) of the WoZ method, and (d) tell us something about what you learned about the HCI of a human-in-the-loop system.  

Please write up a 1-page summary of your experience. Was this a useful exploration for you? What did you learn? What were you trying to find out? How did you set up the study?