Ethics in the Age of Generative AI - Generative AI and Ethics - the Urgency of Now

Source: My personal notes from course Ethics in the Age of Generative AI - Generative AI and Ethics - the Urgency of Now taught by Vilas Dhar, AI Ethicist, on LinkedIn Learning

Introduction and Summary

Like other tools, generative artificial intelligence (AI) is always changing. it is used for good and responsible and negative use cases.

Goal of course is to develop your ethical analysis. Ask who does this affect? support?how does it make us more human?

Ethical Analysis Skill and Framework

Use Cases

Good responsible technology use cases:

• People - help managers with their teams, identify talent
• Insurance - determine when it is needed
• Banking - help when people need financial assistance and services

Negative use cases: deep fakes, inaccurate chatbots, legal issues in AI creations, biased advice like AI used in hiring.

Developing new Frameworks

How can we help people determine AI bias and ethical services?

Use these 3 areas and ask questions.

1. Responsible data practice

   • Ask what is the source of the data?
   • What has been done to reduce bias? historic bias?
   • How do prevent future bias?

2. Boundaries on safe and appropriate use

   • Develop the people to be served and vision the service will do
   • What is the target audience?
   • What are their goals and a responsible way to have them meet the goals?

3. Transparency

   • How did the tool arrive at its output?
   • Fairness testing
   • Can decision makes understand analysis and output?
   • Is the output fair and reviewed by stakeholders?

Example Scenario

An AI chatbot for an organization is making inappropriate and offensive statements to customers. The technologists turns off the chatbot

Data: it came from internet chats, new data will now be from existing customer and company chats and are analyzed and scrubbed

Boundaries: the customers are using the chatbot for use beyond the chatbot’s scope. The chatbot will be restricted to relevant topics and limit non-business conversation

Transparency: input-output checkpoints are put into the chatbox. The organization audits outputs and has a incident management process for issues

Conclusion: ethics must be in the developent process; however, it is never too late to do it.

AI Ethics and Organizations

Ethical Data Organization

Ethical approaches reduce risk and improve outcomes. Ethical data organization involves privacy, reducing bias, and transparency.

Benefits include customer trust. Poor data organization can results in data loss and mistrust of the organization.

Tools:

• Privacy audit: data collection, use and storage
• Training of organization on topics above
• Assess input data to make sure it is inclusive, like culture, people, ability and appropriate to the target audience
• Communicate to all stakeholders like employees, supplies, customers how data is used and how people can access their data

Technology Teams and Ethics

Considerations of people in technology: regulations, specialized skills, environmental impact, deadlines

Tools for creating an ethical culture:

• Ethical and open communication - open to raising issues, discuss impacts
• Training

Executive Communications

Considerations on risks, opportunities. Executive set the culture for AI development and governance.

Executive include corporate officers, organization board directors, and top organization leaders.

1. Tools

   Corporate officers:

   • Responsible AI Policy and Governance Framework - for example policies on data, data validation, design, transparency, training, audits, measurements
   • Person or team responsible for AI development

   Board or top organization leaders:

   • Organizational values
   • Regulatory responsibilities
   • Processes for audits and incidents
   • Resource and people allocation
   • Get and give advice for example an AI committee

Customer Communications

“LISA” Principles based on user research and preferences on businesses:

Listen to users before building

Involve customers in decisions

Share privacy policies

Audit work

If principles are followed, user engagement and satifaction are proven to increase.

Tools:

• User research
• User advisory board
• Privacy policies
• Regular audits by people outside service
• Risk assessments
• Example framework: Google AI Principles, IBM AI Ethics Board that reviews AI projects

Organizational Communications

“ETHICS” communication framework which shows audiences for communications.

• Executives and board members
• Technologists
• Human rights advocates
  • Monitor and advocate for ethical use
• Industry Experts
• Customers
• Society
  • Check impact to society

Tools:

• Communication plan
• Training
• Integrate teams and communications across stakeholders
• User feedback
• Speak with external audiences

Integration of this Learning in Daily Life

Use your expertise in your community, organization

Build skills in technology and social impact

Value stewardship of a human centered future for human excellance