Artificial Intelligence (CS 182) – Fall 2023

Module Topic: AI, Responsibility, and Impact
Module Author: Anni Räty

Course Level: Upper-level undergraduate
AY: 2023-2024

Course Description: “Artificial Intelligence (AI) is already making a powerful impact on modern technology, and is expected to be even more transformative in the near future. The course introduces the ideas and techniques underlying this exciting field, with the goal of teaching students to identify effective representations and approaches for a wide variety of computational tasks. Topics covered in this course are broadly divided into search and planning, optimization and games, and uncertainty and learning. Special attention is given to ethical considerations in AI and to applications that benefit society.” (CS 182 course site.) 

Semesters Taught: Fall 2021, Fall 2022, Fall 2023

    Tags

Module Overview

In this module students practice ethically evaluating the consequences of computing professionals’ choices. The module introduces students to ACM’s code of ethics, and to philosophical concepts of welfare, respect, and justice. The module focuses on a case study where AI is used to optimize NGO interventions to improve maternal and child health outcomes in rural India (SAHELI).

    Connection to Course Technical Material

In this class students learn about topics in artificial intelligence. The SAHELI case study connects with two prior class modules on restless multi-armed bandits (RMAB).

Goals

Module Goals

1. Introduce students to the idea of professional responsibility.
2. Provide students with tools for ethically evaluating the impact of AI systems.
3. Practice identifying interventions into an AI system that affect its ethical impact.

Key Philosophical Questions

1. How can we ethically assess outcomes?
2. How can a computing professional make responsible decisions?
3. How might a system that aims to do good, fall short of norms of justice and respect?

Materials

    Key Philosophical Concepts

• Welfare
• Justice
• Respect
• Responsibility

    Assigned Readings

• Verma, Shresth, Gargi Singh, Aditya Mate, Paritosh Verma, Sruthi Gorantla, Neha Madhiwalla, Aparna Hegde et al. “Increasing impact of mobile health programs: SAHELI for maternal and child care.” In Proceedings of the AAAI Conference on Artificial Intelligence, vol. 37, no. 13, pp. 15594-15602. 2023.

Implementation

Class Agenda

1. Introducing Embedded EthiCS, the ACM code of ethics, and professional responsibility. 
2. Think-Pair-Share: We often have to ask ourselves what would be the responsible thing to do (voting, consumer decisions, vaccinations). Share the last time you had to do this, and how you made a decision.
3. Three kinds of ethical impact:
   1. Welfare
   2. Respect
   3. Justice
4. Case study: Armman and SAHELI. Using RMAB to optimize NGO interventions promoting maternal and child health.
5. Group discussion: Not knowing how SAHELI actually impacted the participants in Armman’s program, how might a system like this…
   1. promote/fail to promote welfare?
   2. promote/fail to promote respect?
   3. promote/fail to promote justice?
6. Discussion of SAHELI’s actual impact. What kinds of changes could be made to the system to improve its impact?

    Sample Class Activity

This module begins with a “Think-Pair-Share” exercise where students identify occasions where they had to decide what would be the responsible thing to do. Later in the module, students discuss the SAHELI case study and its potential ethical impact in small groups (3-4). 

    Module Assignment

Consider the applications for AI that you have encountered in this class. Choose one to evaluate its impact in terms of either welfare, justice, or respect. What changes would improve the impact of the application? For example…

• Changes in design?
• Changes in data?
• Changes in the social context?
• Changes to policy?

Lessons Learned

Student reception was overall positive. The SAHELI case study worked well for this module, but there are other options in the course syllabus as well.
This is a large lecture course where students may not be used to group discussion, so structuring the group discussions and prompting students to share is essential.

Additional Research Notes

• While conducting your research for this module, what materials (articles, blog posts, podcasts, etc.) did you find most helpful? Please include one or two sentences by way of explanation if it won’t be obvious to a future GF what role the material played in your preparation for the module.
• While developing this module, did you have any ideas that were left on the cutting room floor (i.e. ideas about potential topics, readings, activities etc. that were not ultimately incorporated in the module or final repository entry)? If so, please record them here and briefly explain why they did not make the final cut (e.g. time constraints, CS instructor preferences, etc.).
  1. Many of the applications discussed in this course could be used as a case study. CS course heads had recommendations available.
• A more informal take on lessons learned: What else should a future GF know if we have the opportunity to run this module again? For example, based on the actual performance of the module you may have additional insights or speculations to share – If a class activity was successful, do you think the class size was a key factor? Did you perceive any differences between undergraduate vs. graduate students with respect to receptiveness to the module or success on the assignment? Etc.
  1. Class activities went well, but some students required a bit of nudging to join in the conversation. 
  2. The module assignment was very short for this iteration, in part because the module was delivered on the last day of class. There could be a more substantial writing assignment if the module is delivered earlier in the semester.
• After reviewing the student feedback form for the module, were there any comments or general takeaways that you think would be useful for future GFs to take into consideration if they are tasked with repeating this module?