4 An AI Literacy Course for Anyone (Computer Science)
|
|
Description of resource(s):
The website documenting the one-credit CS 109 course “Essentials of AI” offered in Fall of 2023 includes a high-level description of the course, the full reading list, and videos of all the course lectures. Instructors can pattern future courses off of the syllabus. Students can learn directly from the materials.
Links to Resources:
Course webpage with class description, full reading list, and lecture videos that provide an introduction to diverse AI technologies and their ethical and social implications.
Why I implemented this:
The resource provided is the website documenting the lecture videos from the course “CS 109: The Essentials of AI for Life and Society” at the University of Texas at Austin, along with reflections on the experience of running the course that may inform future instructors. The goal of this course was to provide a non-technical overview to any student at the University (or faculty, or staff, or even general public) of the essential knowledge about Artificial Intelligence. Informally, my co-instructors and I considered it a class on “AI literacy.” The first lecture is an overview of the field and where we stand currently. Following that, about ⅔ of the course focuses on the main paradigms and application areas of AI, including planning, probabilistic reasoning, machine learning, computer vision, robotics, natural language processing, and the philosophical and computational foundations of AI. The final ⅓ of the course focuses on the ethical and societal implications of AI, such as potential for mis/disinformation, bias, impacts on the workplace, and possible existential threats.
My main takeaways:
The main takeaway is that there is a very large demand for easy-to-digest resources providing a basic literacy in Artificial Intelligence from a non-technical perspective. Hundreds of students signed up to take the course (for credit or as auditors), and post-class survey responses from students were largely positive, which indicated a desire by many of the students for a more in-depth treatment of the material. Based on the feedback, we have determined that it is worth expanding the course to a full 3-credit-hour course in the coming year.
What else should I consider?
Timing: The time required for that depends on how closely they would like to stick to the curriculum and resources we provide.
Context: This resource is useful for anyone trying to convey an overview of Artificial Intelligence as a discipline to non-experts.
Adaptations: For students, it should be usable/accessible out of the box. For instructors, they could use it as a starting point to adapt it for their own teaching.
Lessons Learned
Below are some lessons learned from teaching the course, and from examining the student feedback collected on surveys throughout and after the course.
Engaging students from a wide variety of backgrounds can pose challenges for an AI literacy course.
Enrollment in the course included 131 students and 584 auditors, which included faculty, staff, postdoctoral fellows, graduate students, and participants external to the University of Texas at Austin (UT). All UT Colleges were represented by auditors who enrolled in the course. Auditors who were not taking the course for credit reported engaging in less of the course and some expressed having different expectations for the course (e.g., wanting to learn how to use AI tools, a greater focus on incorporating AI in teaching).
Course participants appreciated the variety of speakers and connection to examples—the more the better!
22% of the 122 open-ended responses described the variety of guest speakers and broad overview. “I loved that we heard from different guest speakers every week; it was great to get to see so many experts in so many different fields. I also found the course content to be sufficiently reflective of the instructors’ thesis that the AI problem is an interdisciplinary problem, not just a computer science problem as is commonly thought of.” Some also highlighted the value of connections to examples: “Anytime the material was tied into real life examples or scenarios I found it very helpful and applicable. I also liked hearing from speakers with knowledge of different areas of AI.” The lectures stood out as a useful learning tool with 37% of open-ended responses mentioning them: “I liked the lectures and content about how AI works at a fundamental level and what we need to be aware of as we start integrating its use in our real lives.”
The course achieved its main goal of improving AI literacy.
The following questions were administered through the survey, and participants were asked to rate their skill level before the course and now, after taking the course on a 5-point Likert scale where 1 = strongly disagree and 5 = strongly agree.
Questions:
- I can define artificial intelligence (AI).
- I have the necessary vocabulary to discuss AI.
- I can list five examples of AI.
- I can describe how AI affects my daily life.
- My understanding of AI makes me well-equipped to apply it to my future professional work.
- I am prepared to weigh in on the deployment of AI in products that affect me.
- I can evaluate news stories about AI.
- I can differentiate AI science from science fiction.
- I am literate about the technical components of AI.
- I am literate about the societal implications of AI.
Figure. Retrospective pre-/post-survey questions. Each result is statistically significant, with a p value < 0.01, n = 151 for all questions, except for Q5 where n = 150. The survey section prompt read: “For this section of the survey, please reflect on your ability BEFORE you took this course, and compare that to your ability NOW after taking the course. Please indicate the degree to which you disagree or agree with the following statements both BEFORE taking the course and NOW, after taking the course.”
Creating an engaged student community in a one-credit course was challenging.
Some course structure critiques (13%) focused on interactivity and specific assignments: “I would like to see an active student community, either making the lectures way more interactive, or implementing new and fun ideas for students to participate in.”
The readings were the most challenging part of the course for our non-technical audience.
The course was designed for a broad, non-technical audience. However, some still found the level of readings challenging. In response to the open-ended survey item: “What aspect(s) of the course needs improvement?”, most of the content-related comments (19% of 115 open-ended responses) focused on the level and length of the reading: “In my opinion, some of the reading was mathematically challenging, so I would recommend adding prerequisite courses for students.” We provide our reading list on the website, and have flagged some readings that may benefit from additional introductory materials below.
We offer the following suggestions:
- Readings 02, 04, 09, and 10 were described by some students as too technical and/or difficult terminology-wise. We recommend adding introductory pre-reading options for these articles. For some of these readings, a glossary of terms may be helpful for students; for reading 04 a student suggested the inclusion of a key to the mathematical symbols would be helpful to them.
- Course participants expressed a desire for more optional readings on current events. We recommend the addition of informal news articles as supplemental reading wherever possible.
- Students became very engaged when examples connected to their daily lives. To appeal to a broad audience, we recommend ensuring that relatable examples are presented at the beginning of each lecture, and referenced frequently throughout.
As a one-credit-hour course, each week’s readings (and lectures) were fairly independent. While the main course instructors tied things together a little bit with a brief introduction of each speaker and during the Q&A session after each lecture, it would have been helpful to have a 2nd session each week for the instructors to provide more context for each reading and to draw connections between them. As described below, that is part of the plan for the future.
Plans for the future
Based on the feedback from this first version of the course, we intend to expand it to a full 3-credit-hour course, first to be offered in fall of 2024. The expanded course is planned to cover the same set of topics, and will remain geared towards a broad audience, with no technical background required. We acknowledge challenges in satisfying the expectations of students in addition to a broad pool of auditors, and plan mainly to target students in this iteration.
The three-credit course will include an additional synchronous session each week in which the main instructor will provide additional context and lead discussions about the material. In addition to interactive reading assignments and discussion to create an engaged student community, a series of (non-programming-based) assignments will familiarize students with the capabilities and limitations of some AI-based tools and technologies. A significant writing assignment on the societal and ethical implications of AI-based technologies will engage students in considering real-world applications of AI; accordingly, the course will satisfy an ethics requirement for UT Austin.
