Stanford Seminar - A tentative mea culpa - looking back to help the next gen of CSCW researchers

Key Concepts

• Sociotechnical systems
• Downstream consequences
• Wicked problems
• Utopianistic imaginaries
• Governance
• Privacy
• Misinformation
• Expertise sharing
• Surveillance capitalism
• Path dependencies
• Algorithmic auditing
• Regulation
• Emergent network effects

Mea Culpa: A Retrospective Look at CSCW with Implications for Future Research

Introduction

The talk is a personal reflection on the speaker's experiences with CSCW projects, specifically focusing on the unintended negative consequences of well-intentioned work. It explores the challenges of predicting and controlling the downstream effects of technology, using case studies of privacy and misinformation. The speaker acknowledges the positive impacts of CSCW/HCI but focuses on the "less benign" aspects and the "wicked nature" of the problems encountered.

Successes of CSCW/HCI

• Major impact on products like Google Docs, Zoom, distributed games, scientific collaboratories, and the web.
• Facilitated distributed work for the COVID vaccine development, which was described as "this generation's moonshot."
• Intellectual contribution of the concept of technologies being sociotechnical in nature, now recognized even in NIST reports on AI risk.
• Incorporation of new intellectual frameworks like practice theory, postmodernism, and STS.
• Current research tackling serious issues around fairness, accountability, understandability, and transparency.

Case Study 1: Misinformation and Expertise Sharing

• The Initial Vision: The early internet, as envisioned by thinkers like Reinhardt Rangold, was expected to foster understanding and harmony through online communities.
• Expertise Sharing: The speaker's work focused on expertise sharing, exemplified by a University of Washington study where a woman sought advice on teaching while undergoing chemotherapy.
• The Problem: Expertise sharing became a vehicle for misinformation, with statistics showing widespread belief in government conspiracies about vaccines and healthcare.
• Achan Study: The speaker conducted a study of Achan (now 8kun), an extremist online community, observing their efforts to subvert mainstream narratives, including dissecting the Podesta emails.
• Historical Context: The speaker notes that misinformation and conspiracy theories are not new, citing examples like the McCarthy era, the Red Scare, and Henry Ford's anti-Semitism.
• The Shift: Knowledge sharing, initially seen as positive, became a conduit for negative content, hard to handle at scale. The speaker admits to having bought into the utopian vision and not foreseeing the consequences.

Case Study 2: Privacy and Surveillance Capitalism

• P3P Project: The speaker discusses their involvement with the Platform for Privacy Preferences (P3P) project at the World Wide Web Consortium (W3C).
• Goal of P3P: P3P aimed to create a labeling protocol for personal data, allowing users to automatically negotiate privacy agreements with websites through intelligent agents.
• The Social Technical Gap: The speaker references their paper on the social technical gap, arguing that P3P required users to understand an eight-dimensional space, making it impractical.
• Origins of P3P: P3P was created in response to calls for government regulation of the web in the mid-1990s, with the tech industry proposing it as a technical solution to avoid regulation.
• Unintended Consequence: Zuboff argues that P3P inadvertently contributed to surveillance capitalism by capturing data for advertising and other purposes.
• Facebook's Role: Facebook's introduction of customized ads in 2011, driven by the need to attract and retain users, led to the spread of controversy and misinformation for revenue.

Analysis and Lessons Learned

• Underappreciated Problems: The speaker acknowledges that the downstream consequences of these projects were not understood at the time of launch.
• Utopianistic Imaginaries: The speaker argues that utopianistic imaginaries can be used to prevent governance and that technical progress towards these utopias can hinder useful regulation.
• Path Dependencies: The two cases are conjoined, with one leading to the next over time.
• Social Impact: The speaker notes that these issues have harmed social cohesion, social stability, and the sense of collective sociality.
• The Question of Inevitability: The speaker questions whether these negative outcomes were inevitable and what lessons can be learned.

Causes and Challenges

• Difficulty in Seeing Downstream Consequences: It's hard to foresee the emergent network effects that spread exponentially and have societal-level reach.
• Wicked Problems: The world is too complex and dynamic to fully understand the downstream consequences of technology.
• Emergent Properties: New capabilities and uses of technologies like LLMs are constantly being discovered, making it difficult to predict their impact.

Potential Solutions and Future Directions

• Worst-Case Scenarios: The speaker dismisses options like "living with it" or picking commercially uninteresting problems.
• Technical Solutions: The speaker is skeptical of purely mathematical definitions solving fairness and accountability issues.
• Self-Regulation: Self-regulation by industry has historically been ineffective.
• Critical Computing: While acknowledging the value of a leftist critique of neoliberalism, the speaker seeks more immediate and practical solutions.
• Long-Term Governance: The speaker emphasizes the need for long-term governance, which is lacking in the US.
• Regulation: Regulation is often useful for addressing unseen or mispriced externalities, but it is heavily contested and difficult to implement when consequences are unclear.
• Incremental Improvements: The speaker advocates for incremental improvements towards long-term governance.
• Algorithmic Auditing: Pushing for algorithmic auditing to understand what algorithms do and don't do.
• Proactive Proposals: Developing proposals for addressing problems before they arise.
• Background Studies: Conducting background studies for policy people to educate them about the technical and social implications of technology.

Conclusion

• Two Gaps: The speaker identifies two gaps that are difficult to resolve: the gap between what we know now and what might be, and the gap in how to govern or control technology once we understand its effects.
• Federal AI Order: The speaker notes that while there have been some efforts to address AI risks, these efforts are subject to political changes.
• Call for Debate: The speaker concludes by calling for a debate on the possibilities for addressing these challenges.
• The speaker emphasizes that the issues discussed are not just personal but also significant social issues.