AI and Legal Tech in Context: Privacy and Security Commons
- 1. AI and Legal Tech in Context: Governing Privacy and Security Commons How effective privacy and information security depend on formal and informal institutions that encourage sharing knowledge, information, and data. Michael Madison University of Pittsburgh School of Law & Pitt Cyber ACBA February 2018 @profmadison & knowledge-commons.net
- 2. Artificial intelligence / algorithms / cognitive computing raise numerous non-technical public policy questions in legal settings as well as elsewhere Privacy concerns ● security concerns ● transparency concerns ● accountability concerns ● accessibility concerns ● monopoly & competition concerns ● cooperation & exclusivity concerns ● “nature of humanity” concerns Normative / conceptual / theoretical questions: What role{s} should AI / algorithms play? Descriptive / research-based questions: How do systems for generating, distributing, accessing, and managing information operate in practice?
- 3. The research question How is effective privacy / security accomplished? The hypothesis: governance Privacy and security are grounded in commons institutions (i.e., they depend on formal and informal group-based patterns of practice and belief) and require more than technical solutions (e.g., authentication techniques) or legal rules (e.g., criminal enforcement). @profmadison & knowledge-commons.net
- 4. The research strategy We study information and knowledge commons, institutions that generate and protect information by sharing it in the context of rule-based systems. Commons are Institutionalized sharing of resources among members of some group or community, solving some social dilemma. Not a place. Not a thing. Not “the commons.” @profmadison & knowledge-commons.net
- 5. 5 The research to date • Defined the Knowledge Commons Research Framework (Madison et al. 2010) (building on Ostrom 1990) • Governing Knowledge Commons (Frischmann et al. Oxford UP 2014) • Governing Medical Knowledge Commons (Strandburg et al. Cambridge UP 2017) • Governing Privacy Commons (forthcoming Cambridge UP) • Governing University Commons (forthcoming Cambridge UP) @profmadison & knowledge-commons.net
- 6. Privacy and security commons research in progress In case study context, identify the informational dilemma(s) to be solved; the resources to be managed; the group or community; the formal and informal rules by which information in the group or community is governed (internal and external); the outcomes – good or bad. Privacy/security examples implicate (i) mix of privacy & collaboration rules governing information resources (ii) in order to promote valuable practices: • Anonymous, private, secure voting systems encourage democratic participation and lead to aggregating political preferences in fair ways • Financial institutions’ secure collection & management of customer data encourages participation in financial markets • Secure sharing of private information in social networks (both close-knit & loose-knit) can promote healthy community and society • Chatham House Rule for confidential meetings encourages productive collaboration @profmadison & knowledge-commons.net
- 7. Strengths and weaknesses • Contextual approach leads to learning more about variance in communities, obstacles/dilemmas, objectives, and institutions beyond tech firms, beyond markets, beyond governments • Overlaps and intersections among commons institutions can be explored • Bottom up learning about normative values • Possibility of improving institutional design via design principles BUT • The approach doesn’t work at the extremes: privacy with n=1; privacy with n=everyone • The approach is complicated by working with physical / material resources • Sidelines normative debate and values • Essentially ethnographic; needs dedicated research community @profmadison & knowledge-commons.net
- 8. Payoffs: Artificial intelligence, algorithms, and cognitive computing in context • Commons governance – sharing information resources to generate productive outcomes – is historical, traditional, and effective. • Distinguish between information system as infrastructure (single resource, multiple uses & users) and system as proprietary service (an exclusive thing). Commons is more likely to be effective as governance for infrastructure. • Payoff 1: Exclusivity matters most when AI is used in consulting one-to-one with clients and delivering services to clients. Example: predictive analytics. • Payoff 2: Commons may matter more, and may trump exclusivity, where legal tech / AI operates as infrastructure – e.g., resource in advocacy and/or judicial administration. Example: DNA testing, election security. Similar: case text databases. • Payoff 3: Governance cannot be divorced from hard values questions. Capabilities of AI may challenge distinctions between humans and tech in framing big “what is justice?” questions. If AI can write briefs (as Westlaw can, or will soon), and if AI can adjudicate disputes (as insurance carriers may soon do, with simple claims), then why have lawyers? That’s not a rhetorical question.
- 9. Thank you! @profmadison & knowledge-commons.net