1 of 19

Ethics of �Data and AI

September 29, 2023

100 Holliday Street

Baltimore, MD 21202

​

Prepared by BDL on

​

A Community for Learners

Brought to you by

2 of 19

Our disciplines

New look. Same passion for civic design.

Ethics of

Data and AI

2

Program Management

Human-Centered Design

Data Science

3 of 19

Videos, links, and more now online.

Visit our BDL page at�opi.baltimorecity.gov/baltimore-design-lab

3

View our past sessions.

4 of 19

Icebreaker

Ethics of

Data and AI

Describe a personalized recommendation from a company (Netflix, Amazon, etc.) that was either a) uncannily accurate or b) out of the blue!

Put your answer in the chat!

4

5 of 19

Welcome to today’s facilitator

Ethics of

Data and AI

5

Justin Elszasz

Chief Data Officer

City of Baltimore

6 of 19

Agenda

Ethics of

Data and AI

6

What is AI?

(5 min)

What are the concerns?

(10 min)

How might local government use AI?

(5 min)

​

1

2

3

7 of 19

Ethics of

Data and AI

7

“We cannot be trusted with the stuff that we come up with. The machinery could eat us we just really love our buttons.”

​

Aesop Rock

​

​

8 of 19

What is Artificial Intelligence?

A computer program that can be trained using data to complete a task.

​

Ethics of

Data and AI

8

Input

Output

AI Algorithm​

9 of 19

Ethics of

Data and AI

9

Photo or Video of a Person

AI Algorithm​

Name of Person

Vehicle Cameras/ Sensors

AI Algorithm​

Vehicle Direction

Homebuyer Credit History, Demographics

AI Algorithm​

Likelihood of Loan Default

10 of 19

Ethics of

Data and AI

What is Artificial Intelligence?

  • Find a needle in a haystack
  • Flag stuff early
  • Prioritize a backlog
  • Optimize Resources
  • A/B Test
  • Generate Content �Need to draft an email? A report? Edit video? Create a brand new image? This is generative AI.

​

10

AI can solve several types of problems relevant for cities

11 of 19

What is different about AI? �Why should we care?

  1. AI can make a large number of decisions quickly and accountability is murky.
  2. Because AI is built using historical data, existing inequities and power structures can not only be perpetuated, but worsened.
  3. AI-generated content muddies the waters around authorship, authority, and reality.

Ethics of

Data and AI

11

12 of 19

Should we be scared yet?

AI currently can only perform the narrow set of tasks for which it is trained.

It can not learn to perform new, previously unseen tasks (this is called Artificial General Intelligence, or AGI.)

Ethics of

Data and AI

12

13 of 19

What are some concerns?

Ethics of

Data and AI

13

Privacy

Bias

Explainability

Labor

14 of 19

Concerns | Privacy

Ethics of

Data and AI

Privacy is comprised of:

  • Transparency�What is being collected?
  • Control�Where is the data going and can it be controlled?
  • Full Services�If a user opts out, do they lose access to services?
  • Opt In/Out�Can a user opt out?

14

Cookies store information about you on the website. Depending on the website's privacy policy, part of the ‘cookie’ may be shared with other companies, for example, to make targeted ads or connect platforms, like connecting your Instagram and Facebook.

15 of 19

2016 ProPublica investigation of criminal sentencing risk scores found:

“The formula was particularly likely to falsely flag black defendants as future criminals, wrongly labeling them this way at almost twice the rate as white defendants.”

“White defendants were mislabeled as low risk more often than black defendants.”

Explicit variables for race, gender, etc. don’t need to be included for bias to propagate.

​

Concerns | Bias

Ethics of

Data and AI

15

16 of 19

Concerns | Explainability

Ethics of

Data and AI

Explainability is comprised of:

  • Informed Consent�What constitutes being informed of how your data are used?
  • Black Box�AI recognizes patterns not discernible by humans.
  • Global Explainability�An explanation of how a model makes its predictions in general.
  • Local Explainability�An explanation of why any one particular prediction was made.

16

Input

Output

???????????????????????????​

17 of 19

Concerns | Labor

Labor issues include:

  1. We built AI�Our data has been used to train these tools. What are we owed?
  2. “Ghost labor”�Workers in middle income countries are often employed to label data to feed into models for low wages.
  3. Automation�Many workers in manual labor fear that AI will automate their jobs out of existence. How do we mitigate this concern?

17

Read “Atlas of AI” by Kate Crawford for more in depth treatment of labor issues.

18 of 19

How might �local government use AI?

Ethics of

Data and AI

Let’s imagine –

18

19 of 19

Questions?

Ethics of

Data and AI

19