ai policy - about - sigai · 2020. 8. 12. · ai matters, volume 3, issue 4winter 2018 systems, and...

AI MATTERS, VOLUME 3, ISSUE 4 WINTER 2018

AI PolicyLarry Medsker (George Washington University; lrm@gwu.edu)DOI: 10.1145/3175502.3175510

Abstract

AI Policy is a regular column in AI Mattersfeaturing summaries and commentary basedon postings that appear twice a month in theAI Matters blog (https://sigai.acm.org/aimatters/blog/). Selected posts are summarized in is-sues of AI Matters.

Introduction

The SIGAI Public Policy goals are to:

• promote discussion of policies related to AIthrough posts in the AI Matters blog on the1st and 15th of each month,

• help identify external groups with commoninterests in AI Public Policy,

• encourage SIGAI members to partner inpolicy initiatives with these organizations,and

• disseminate public policy ideas to the SIGAImembership through articles in the newslet-ter.

I welcome everyone to make blog commentsso we can develop a rich knowledge base ofinformation and ideas representing the SIGAImembers.

Organizations Related to AI and Policy

In 2017 we expanded our reporting on andwork with the American Association for the Ad-vancement of Science, particularly the Cen-ter of Science, Policy, and Society. WhileAAAS policy issues are usually not directly re-lated to AI, a regular look at their Policy Alertnotifications is useful for larger policy issues,and helpful to see opportunities for SIGAI tobe involved in public policy events. We alsoexpanded our relationship with the ACM USPublic Policy Council (USACM), which servesas the focal point for ACM’s interaction withUS government organizations, the computing

Copyright c© 2018 by the author(s).

community, and the US public in all matters ofUS public policy related to information technol-ogy. USACM addresses issues in innovation,privacy, security, digital governance, intellec-tual property, accessibility, and e-voting. I ama member of USACM, in part representing ourACM SIGAI.

Algorithmic Transparency andAccountability

ATA is a current major initiative with USACM.Your Public Policy Officer attended and re-ported on the USACM Panel on AlgorithmicTransparency and Accountability, ehcih tookplace on Thursday, September 14th at theNational Press Club. The panelists weremoderator Simson Garfinkel, Jeanna NeefeMatthews, Nicholas Diakopoulos, Dan Rubins,Geoff Cohen, and Ansgar Koene. USACMChair Stuart Shapiro opened the event, andBen Sneiderman provided comments from theaudience.

USACM and EUACM have identified and cod-ified a set of principles intended to ensure fair-ness in this evolving policy and technologyecosystem. These were a focus of the paneldiscussion and are as follows: (1) awareness;(2) access and redress; (3) accountability; (4)explanation; (5) data provenance; (6) audit-ability; and (7) validation and testing. See alsothe full letter in the September 2017 issue ofCACM.

The panel and audience discussion rangedfrom frameworks for evaluating algorithmsand creating policy for fairness to examples ofalgorithmic abuse. Language for clear com-munication with the public and policymakers,as well as even scientists, was a concern—ashas been discussed in our Public Policy blog.Algorithms in the strict sense may not alwaysbe the issue, but rather the data used tobuild and train a system, especially when thesystem is used for prediction and decisionmaking. Much was said about the types ofbias and unfairness that can be embeddedin modern AI and machine learning systems.

32

AI MATTERS, VOLUME 3, ISSUE 4 WINTER 2018

The essence of the concerns includes theability to explain how a system works, theneed to develop models of algorithmic trans-parency, and how policy or an independentclearinghouse might identify fair and prob-lematic algorithmic systems. Please readmore about the panel discussion at https://www.acm.org/public-policy/algorithmic-paneland view the video at https://www.youtube.com/watch?v=DDW-nM8idgg&feature=youtu.be.

AAAI Fall 2017 Symposium Series

This year’s Fall Symposium Series (Novem-ber 9-11) https://aaai.org/Symposia/Fall/fss17symposia.php provided updates andinsights on advances in research and tech-nology, including resources for discussion ofAI policy issues. The symposia addressedtopics in human-robot interaction, cognitiveassistance in government and public sectors,military applications, human-robot collabo-ration, and a standard model of the mind.Important themes for public policy were aboutthe advances and questions on human-AIcollaboration.

The cognitive assistance sessions this yearfocused on government and public sectorapplications, particularly autonomous sys-tems, healthcare, and education. Human-technology collaboration advances involveddiscussions of issues relevant to public pol-icy, including privacy and algorithmic trans-parency. The increasing mix of AI with hu-mans in ubiquitous public and private sys-tems was the subject of discussions aboutnew technological developments and the needfor understanding and anticipating challengesfor communication and collaboration. Particu-lar issues were on jobs and de-skilling of theworkforce, credit and blame when AI applica-tions work or fail, and the role of humans withautonomous systems.

IBM’s Jim Spohrer made an outstanding pre-sentation “A Look Toward the Future,” incor-porating his rich experience and current workon anticipated impacts of new technology. Hisslides are well worth studying, especially forthe role of hardware in game-changing tech-nologies with likely milestones every ten yearsthrough 2045. Radical developments in tech-nology would challenge public policy in waysthat are difficult to imagine, but current poli-

cymakers and the AI community need to try.See related references in the Resources sec-tion below.

Particular takeaways, and anticipated subjectsfor future blogs, are about the importance oflikely far-reaching research and applicationson public policy. The degree and nature ofcognitive collaboration with machines, the fu-ture of jobs, new demands on educational sys-tems as cognitive assistance becomes deepand pervasive, and the anticipated radicalchanges in AI capabilities put the challengesto public policy in a new perspective. AI re-searchers and developers need to partner withsocial scientists to anticipate communicationand societal issues as human-machine collab-oration accelerates, both in system develop-ment teams and in the new workforce.

Collaborations with other Policy andEthics Groups

SIGAI recently began discussion with othergroups, particularly between ACM and IEEE,on finding ways to bring together efforts onAlgorithmic Transparency and Accountability.One opportunity is at RightsCon Toronto: May16-18, 2018. The call for proposals mentioned“Artificial Intelligence, Automation, and Algo-rithmic Accountability’ as one of their program“buckets.”

As AI is becoming more pervasive in our lives,the impact on society is increasingly signifi-cant. Concerns and issues are being raisedregarding value alignment, data bias and datapolicy, regulations, and workforce displace-ment. We need multi-disciplinary and multi-stakeholder efforts to find the best ways to ad-dress concerns using expertise from AI, com-puter science, ethics, philosophy, economics,sociology, psychology, law, history, and poli-tics. AAAI and ACM are joining forces to starta new conference, the AAAI/ACM Conferenceon AI, Ethics, and Society. The first edition ofthis conference http://www.aies-conference.comwill be co-located with AAAI-18 on February2-3, 2018, in New Orleans. The program ofthe conference will include peer-reviewed pa-per presentations, invited talks, panels, andworking sessions. The conference will covera broad set of topics, including trust and ex-planations in AI systems, fairness and trans-parency, ethical design and development of AI

33

AI MATTERS, VOLUME 3, ISSUE 4 WINTER 2018

systems, and impact of AI on the workforce.

Upcoming

Some themes planned for the SIGAI Pub-lic Policy posts for 2018 include algorithmicaccountability, human-machine collaboration,and the impacts of AI and Data Science onthe future of education and the labor market.We will look at potential policies for today thatcould mitigate impacts of AI on individuals andsociety. Policy areas include innovative edu-cational systems, ideas for alternate economicsystems, and regulatory changes to promotesafe and fair technological innovation.

Resources• AAAI information related to science policy

issues:https://aitopics.org/search

• AAAI Symposium Series:https://aaai.org/Symposia/Fall/fss17.php

• Ansgar Koene:https://theconversation.com/machine-gaydar-ai-is-reinforcing-stereotypes-that-liberal-societies-are-trying-to-get-rid-of-83837

• CACM letter on algorithmic transparencyand accountability:https://cacm.acm.org/magazines/2017/9/220423-toward-algorithmic-transparency-and-accountability/fulltext#FNA

• Jim Sphorer, A Look Toward the Future:https://www.slideshare.net/spohrer

• Humans, robotics, and the future of manu-facturing:https://www.engadget.com/2017/09/11/human-robot-ai-manufacturing/

• New education systems and the future ofwork:https://www.edweek.org/ew/articles/2017/09/27/the-future-of-work-is-uncertain-schools.html

• Noriko Arai’s TED talk on Todai Robot:https://www.ted.com/talks/noriko arai can arobot pass a university entrance exam

• Smart phone app “Seeing AI”:https://www.youtube.com/watch?v=bqeQByqf f8

Larry Medsker is Re-search Professor ofPhysics and Directorof the Data Sciencegraduate program atThe George WashingtonUniversity. Dr. Medskeris a former Dean of theSiena College School ofScience, and Professor

in Computer Science and in Physics, wherehe was a co-founder of the Siena Institutefor Artificial Intelligence. His research andteaching continues at GW on the nature ofhumans and machines and the impacts of AIon society and policya. Professor Medskersresearch in AI includes work on artificial neu-ral networks and hybrid intelligent systems.He is the Public Policy Officer for the ACMSIGAI.

ahttp://www.humai.org/humai/ and http://humac-web.org/

34