ethics of algorithms - knowitevent.s3.amazonaws.como… · algorithms, ethics and recruiting loren...

Post on 07-Jul-2020

2 Views

Category:

Documents

0 Downloads

Preview:

Click to see full reader

TRANSCRIPT

Ethics of Algorithms

Dominique Shelton Leipzig

Partner, Privacy & Security, Co-Chair Global Ad Tech & Data Management

Perkins Coie

DSheltonLeipzig@perkinscoie.com

Artificial Intelligence and Privacy What do new privacy laws mean for your AI business?

Part 1: How Did We Get Here?

Part 2: CCPA and AI

Part 3: AI in a Pandemic

2

INTRODUCTION

CCPA: HOW DID WE GET HERE?

DOMINIQUE SHELTON LEIPZIG

3

Global Privacy Trends

India may be next...

4

The GDPR: Rights & Obligations

1 2

3

4 5

6

7

8

Right to Access (can be made irrespective

of lawful basis of processing (Art.15)

Correction Request Right

(Art. 16)

Erasure Requests (Art. 17) -

only valid under six

circumstances)

Restriction Request

(Art. 18)

Right of Portability

(Art. 20; only for consent or

contractual necessity)

Objection Rights

(Art. 21)

Objection to Automated

Processing (Art. 22)

Transparency Obligation

(Art. 13)

5

AI and GDPR

• Valid Legal Basis for

Processing Art. 6

• Human review of

contested automated

decision-making Art. 22

• Deletion

• Correction

• Access

6

New State Proposed Laws and Two New Federal Bills

1. Colorado

2. Connecticut

3. Hawaii

4. Illinois

5. Maryland

6. Massachusetts

7. Mississippi

8. New Jersey

9. New York

10. New Mexico

11. North Dakota

12. Pennsylvania

13. Rhode Island

14. Texas

15. Vermont

16. Virginia

17. Washington No federal preemption in place.

7

2018 and 2019 Ballot Initiatives

• June 2018

Ballot Initiative

Qualifies

• Compromise

Legislation

• CCPA 2.0

11/13/2019

• Federal

Privacy Law

8

On the Horizon in California 2020

• Proposed Regs 10/10/2019; AG says minimal changes expected

• Final CPRA version submitted on 11/13/2019; currently out for signature

9

The CCPA: Rights & Obligations

1 2

3

4 5

6

7

8

Abbreviated Right to

Know Re PI Collection

Expanded Right to Know

RE PI Collection

Detailed Right to Know PI

Sales and/or Disclosures for

a Business Purpose

Right to Opt Out of PI

Sales for Adults

Right to Opt-In to PI Sales

for Kids

Right to Access

and Portability

Right to Deletion

Right Not to Be

Discriminated Against for

Asserting Rights 1-7

10

Three Independent Business Obligations

Targeted

Training

Create Designated Methods

for Asserting Rights

Obtain Immunity by

Making Contract Meet

Specific Criteria

11

CCPA – Artificial Intelligence

12

AI and the CCPA

• Consumer Right Nos. 1-3 Require transparency about

– Specific pieces of PI

– Purposes of Collection

– Sharing

– Sale

• Deletion Rights

• Access Rights

13

AI vs. Privacy to fight the Pandemic

14

Legislative Concerns re: Covid-19 and Privacy

15

CCPA Definition of PI Is Broad

• Identifiers

• Geolocation

• Professional or

employment

information

• Education

information

• Protected

Characteristics

• Commercial

Information

• Internet or Other

Electronic Network

Activity

Information

• Inferences

• Biometric

Information (e.g.,

facial recognition,

fingerprint, voice

recordings,

iris/retina images)

• Audio, electronic,

visual, thermal,

olfactory, or similar

information

16

Perkins Coie Comments to the CA AG on Behalf of

California Chamber of Commerce

CAL CHAMBER COMMENTS

REGARDING REGULATIONS

17

CCPA in the Press

18

Coordinate Responses For Consumer Requests

19

Implementing the CCPA:

A Guide for Global Business

https://iapp.org/store/books/a191P000003QmX3QAK/

20

Speaker Info

DOMINIQUE SHELTON LEIPZIG

Partner and Co-Chair, Ad Tech Privacy & Data Management Practice, Perkins Coie DSheltonLeipzig@perkinscoie.com

21

Algorithms, Ethics and recruiting

Loren Larsen

CTO

HireVue

RESUME

ANALYTICS

TOP CANDIDATE DISTRIBUTION

WITHOUT ASSESSMENTS TOP CANDIDATE DISTRIBUTION

WITH ASSESSMENTS

Simple Candidate Prioritization

ADVERSE IMPACT (4/5THS RULE) ILLUSTRATION

MALE

FEMALE

KEY: Adverse Impact Ratio Calculations

Before Mitigation (33% cutoff score):

Male Passing Rate (MPR): 80% Female

Passing Rate (FPR): 53%

Female Adverse Impact Ratio:

.67 (FPR / MPR)

33%

R VALUE = .42

66%

The Dilemma of Ethics and Algorithms:

Transparency, Secrecy, Innovation, and

Consumer Rights

Samir Mehta

Partner

Stinson, LLP

The Dilemma

• Consumers want:

– To know what data is used

– To know how that data is used

• Businesses want:

– To create new tools that use data in new ways

– To protect their new tools from disclosure

Why Keep Algorithms Secret?

• Short windows to capitalize

– Upfront R&D costs

– Technology gains will erode

• Keeping the edge is crucial

– Secrecy

– Intellectual Property

Concerns of Keeping Algorithms Secret

• Consumer impact is unclear

• Data models are unclear

• Technology platforms

– A common approach can

spread rapidly

• Latency of spotting issues

Solving the Dilemma

• Regulation and legislation

• Industry standards

• Engineering ethics

• Standards

• Watchdogs

• Litigation

• Consumer activism

Each tool is valuable, but each tool is limited

Seeing Past the Dilemma

• Algorithms can provide ethical gains

• Algorithms present opportunity for all parties

– Disrupt problematic past practices

– Reduced cost for solutions

– Free customers to work on more interesting problems

– More insight and accountability

– Speed

– Efficiency

– Government accountability

– Better industry standards

THANK YOU!

35

top related