This video discusses the critical challenge of understanding how millions of people use AI assistants daily while simultaneously protecting their privacy. Although companies collect vast amounts of data from these interactions, public access to this information remains extremely limited, creating a “data gap” that hinders research and safe AI development. To address this, a system called Clio was developed—a privacy-preserving analytical pipeline designed to extract meaningful usage patterns from millions of AI conversations without exposing any individual’s private information. Highlights 🔍 Clio enables large-scale analysis of AI assistant usage without compromising privacy. 🛡️ Privacy is protected through multi-layered automated defenses and aggregation techniques. 👩‍💻 Dominant AI use cases include coding, writing, research, and learning. 🌐 AI usage patterns differ across cultures and languages, reflecting diverse needs. 🚨 Clio detects misuse and abuse patterns early, supporting proactive safety measures. 🤖 Automation minimizes human exposure to sensitive conversations. 🔄 Continuous auditing and monitoring keep privacy protections robust over time. Key Insights 🔐 Privacy by Design is Essential: Clio exemplifies the necessity of integrating privacy into AI system design from day one. By focusing on data minimization, automation, and layered privacy protections, it avoids the pitfalls of ad hoc, reactive privacy measures. This approach is critical for maintaining user trust and meeting ethical and legal obligations. Without such design principles, AI developers risk exposing sensitive user data or failing to learn effectively from real-world usage. 🧩 Aggregation Prevents Re-Identification: Clio’s requirement that clusters represent large groups of conversations ensures individual users cannot be singled out. This statistical barrier is a powerful privacy tool, as it makes linking any data point back to a particular person practically impossible. It balances the need for granular insights with the imperative to protect personal identities, a challenge often underestimated in AI analytics. 🤖 Automation Reduces Human Risk and Cost: By automating fact extraction, clustering, summarization, and privacy auditing, Clio avoids exposing human reviewers to sensitive content. This not only protects reviewer mental well-being but also improves scalability and speed, enabling real-time analysis of millions of conversations. Automation is a key enabler for ethically and efficiently managing large-scale AI data. 🌍 Cultural and Linguistic Variations Matter: The system’s ability to identify different usage patterns across languages and cultures highlights AI’s adaptability and diverse applications. For example, Japanese users discuss elder care more frequently, while Spanish users focus more on finance. Understanding these nuances helps tailor AI development to meet specific community needs and supports more inclusive, globally relevant AI tools. 🚨 Early Detection of Misuse Enables Proactive Safety: Clio’s clustering approach allows safety teams to observe widespread patterns of abuse or harmful behavior without reading individual chats. This macro-level visibility empowers teams to swiftly update safety protocols and filters before issues escalate, shifting safety work from reactive incident response to proactive risk management. 📊 High-Level Insights Replace Guesswork: The ability to generate synthetic summaries that accurately reflect user behavior without revealing personal data transforms how AI developers understand user needs and challenges. This evidence-based approach fosters better product decisions, more effective improvements, and safer AI deployments, moving beyond assumptions to grounded knowledge. 🔄 Ongoing Auditing Maintains Privacy Integrity: Clio’s continuous testing, including red team attacks and privacy audits, ensures that its layered defenses remain effective as AI and user behaviors evolve. This commitment to vigilance is vital, as static privacy solutions can degrade over time. Persistent oversight strengthens the system’s resilience, safeguarding privacy in an ever-changing technological landscape. If you found this useful, please like and share! #AI #Privacy #AISafety #Clio #Anthropic #Claude #Governance

Abstract: This paper presents Clio (Claude insights and observations), a privacy-preserving platform that uses AI assistants to analyze and surface aggregated usage patterns across millions of conversations without requiring human reviewers to read raw user data. The system addresses a critical gap in understanding how AI assistants are used in practice while maintaining robust privacy protections through multiple layers of safeguards. We validate Clio's accuracy through extensive evaluations, demonstrating 94% accuracy in reconstructing ground-truth topic distributions and achieving undetectable levels of private information in final outputs through empirical privacy auditing. Applied to one million Claude.ai conversations, Clio reveals that coding, writing, and research tasks dominate usage, with significant cross-language variations—for example, Japanese conversations discuss elder care at higher rates than other languages. We demonstrate Clio's utility for safety purposes by identifying coordinated abuse attempts, monitoring for unknown risks during high-stakes periods like capability launches and elections, and improving existing safety classifiers. By enabling scalable analysis of real-world AI usage while preserving privacy, Clio provides an empirical foundation for AI safety and governance.

This video explores the profound impact of artificial intelligence (AI) on the modern workplace through a detailed survey conducted using a novel AI interviewing tool called the Anthropomorphic Interviewer. This tool engaged 1,250 professionals across three groups—general workers, scientists, and creatives—to understand how AI is integrated into daily work life, balancing productivity gains with psychological and professional concerns. The findings reveal a dual narrative: AI as a powerful enhancer of efficiency and productivity, and simultaneously a source of anxiety surrounding job security, professional identity, and ethical dilemmas. While the majority report significant time savings and workflow improvements, many workers secretly use AI due to stigma and fears of being perceived as less capable. Scientists emphasize the critical need for verifiable, transparent AI outputs to maintain trust in their work. Highlights 🤖 AI boosts productivity dramatically: 86% of general workers and 97% of creatives report significant time savings. 🧠 Scientists rely on AI for brainstorming but demand transparency and verification of AI outputs. 🤫 69% of workers hide their AI use due to fear of judgment and stigma, especially among creatives. ⚖️ Two AI usage modes: augmentation (AI assists, humans lead) and automation (AI takes over tasks). 😟 Over half of workers feel anxious about their future job security with AI’s rise. 💬 Organizations should promote open dialogue, clear AI policies, and employee involvement to reduce anxiety. 🎓 Training on AI literacy and verification is crucial to build trust and effective human-AI collaboration. Key Insights 🤖 AI as a Productivity Multiplier: The survey underscores AI’s transformative impact on workplace efficiency. With nearly nine out of ten general workers and almost all creatives reporting substantial time savings, AI is reshaping workflows across industries. This shift enables professionals to focus on higher-level tasks, creativity, and strategy rather than repetitive, time-consuming work, marking a paradigm shift in how work is performed. 🧬 The Scientist’s Dilemma: Trust and Verification: Scientists’ cautious embrace of AI reveals a critical boundary for AI adoption—verifiability. In fields grounded in evidence and reproducibility, AI-generated insights must be transparent and traceable to maintain scientific integrity. This need for “explainability” highlights a key challenge in AI development and deployment: without clear audit trails and source validation, AI risks undermining trust and the core values of scientific inquiry. 🤫 Stigma and Secret Usage: The hidden use of AI by 69% of workers reflects a cultural and social barrier to AI acceptance in the workplace. The perception of AI use as “cheating” or a threat to personal skill fosters secrecy and inhibits open learning and collaboration. This stigma is particularly acute among creatives, for whom AI-assisted work raises deep questions about authorship, originality, and professional identity. ⚖️ Augmentation vs. Automation—A Spectrum of Control: The distinction between AI augmentation and automation is crucial for understanding the evolving nature of work. Augmentation positions AI as a collaborative partner enhancing human capabilities without replacing the individual’s creativity or judgment. Automation, by contrast, can shift roles towards oversight and management of AI systems, potentially eroding the sense of agency and fulfillment derived from hands-on work. 😟 Anxiety and Identity Crisis Amid AI Advances: More than half of the surveyed workers express anxiety about their professional futures, highlighting a widespread emotional and psychological impact beyond productivity metrics. The anxiety stems not only from fears of job loss but also from concerns about diminished professional pride and a loss of purpose as AI takes on more tasks. 💬 Leadership’s Role in Shaping AI Adoption: The video stresses that organizational success with AI depends heavily on leadership practices. Transparent communication about AI use, clear guidelines defining acceptable applications, and creating safe spaces for employees to share experiences and concerns are foundational steps. Without this, stigma and mistrust undermine AI adoption and employee morale. 🎓 The Imperative of AI Literacy and Training: Effective integration of AI requires more than just deploying tools—it demands equipping workers with skills to critically assess AI outputs, identify errors, and collaborate effectively with AI systems. Training programs should emphasize verification techniques, error spotting, and understanding AI’s limitations to prevent blind trust and misuse. #AI #Workplace #Anthropic #QualitativeResearch #FutureOfWork OUTLINE: 00:00:00 - Introducing the Study 00:01:36 - Gains and Anxieties 00:03:27 - Scientists and Verification 00:04:45 - How Work Is Really Changing 00:06:11 - Steps for a Kinder AI Future

Abstract: This research introduces Anthropic Interviewer, an AI-powered tool designed to conduct large-scale qualitative interviews at unprecedented scale while maintaining conversational depth. To validate this methodology, we deployed the system to interview 1,250 professionals—comprising 1,000 general workforce participants, 125 scientists, and 125 creative professionals—about their experiences integrating AI into their work. Results indicate predominantly positive sentiment regarding AI's productivity impact, with 86% of general workforce participants reporting time savings and 97% of creatives noting efficiency gains. However, significant concerns emerged around social stigma (69% of general workforce), professional displacement (55% expressing anxiety), and verification reliability (particularly among scientists). Thematic analysis revealed divergent adoption patterns: general workforce professionals envision AI-augmented supervisory roles; creatives navigate productivity gains against peer judgment and identity concerns; scientists desire AI partnership but withhold trust for core research tasks. This study demonstrates both the viability of AI-mediated qualitative research at scale and provides empirical insight into how professionals across diverse domains are experiencing AI's integration into knowledge work.

In this HCI Webinar, Dr. Jonathan H. Westover talks with performance management, workplace dynamics, and employer liabilities. Mark F. Kluger practices exclusively in the area of labor and employment law on behalf of employers. For ten years, before founding Kluger Healey, LLC, he was Chairman of the Labor and Employment Department of one of New Jersey’s oldest law firms. Mark is a frequent speaker and writer on sexual harassment and discrimination avoidance, workplace diversity, performance management, union avoidance, and a myriad of other employment-related subjects and regularly conducts training sessions for employers on these critical topics. In addition, Mark has extensive experience in counseling employers on issues involving discipline and discharge, reductions in force, mergers and acquisitions, compliance with wage and hour, disability, COBRA, and family and medical leave laws. He regularly drafts all forms of employment policies and handbooks, severance agreements, employment contracts, non-competition and confidentiality agreements, and affirmative action plans. Mark also represents employers in collective bargaining, grievance arbitration, NLRB proceedings, and picket line issues. Mark graduated from Vassar College in 1984 and Cornell University Law School in 1987. He was an Adjunct Professor at Seton Hall Law School from 1991-1996 and served as a member and President of the Board of Education in North Caldwell, New Jersey from 2002-2008.

The video transcript explores the evolving preferences of younger workers, particularly millennials and Gen Z, who favor a hybrid work model blending remote and in-office work. This preference is not a rejection of company culture or commitment but a pragmatic response to the demands of modern work. Hybrid work enables employees to engage in deep, focused tasks at home while reserving office days for collaboration, social learning, and spontaneous interactions that foster innovation and relationships. Organizations that rigidly enforce full-time office attendance risk losing emerging talent and limiting their access to a global talent pool, which in turn stifles innovation and diversity. Highlights 🌍 Younger workers prefer hybrid work for a balance of remote focus and in-office collaboration. 🏢 Rigid full-time office mandates risk losing top talent and limit access to global candidates. 💡 Hybrid work challenges stem from outdated leadership, not employee productivity. 🤝 Two-tiered systems between remote and in-office workers damage trust and morale. 🎯 Effective hybrid leadership requires outcome-based management and clear communication. 🛠️ Offices should be redesigned with hybrid work in mind, including tech for remote inclusion. 📈 Fair, location-agnostic performance reviews and mentorship foster growth for all employees. Key Insights 🌐 Hybrid work is a strategic imperative in a globalized talent market: Companies that fail to offer flexible work options limit their talent pool geographically and risk losing innovative, diverse talent essential for competitive advantage. In a world increasingly connected by technology, geographic boundaries should no longer restrict hiring or collaboration. 🔄 Leadership adaptation is critical to hybrid success: Many leadership failures arise from attempting to apply outdated, office-centric management styles to a distributed workforce. The inability to trust remote employees or to measure performance by outcomes rather than physical presence creates inefficiencies and unfair disparities. ⚖️ Creating equity between remote and in-office workers prevents organizational division: When in-office staff receive more visibility and opportunities, remote workers feel marginalized, which erodes trust and damages team cohesion. Intentional structures and policies must ensure equal access to career development, recognition, and social connection regardless of location. 📅 Intentional office scheduling enhances the value of in-person work: Rather than daily attendance, designating anchor days for team-based, collaborative activities maximizes the benefits of physical presence. This approach respects employees’ need for focused remote work while capitalizing on the unique advantages of face-to-face interaction. 🧑‍🏫 Manager training and psychological safety are foundational: Effective hybrid management requires new skills in remote communication, trust-building, and outcome-focused evaluation. Investing in leadership development and fostering inclusive meeting practices create environments where all employees feel valued and engaged. 🖥️ Redesigning office spaces for hybrid work supports productivity and inclusion: Offices should be adapted with technology like speaker tracking cameras and dedicated microphones to integrate remote participants fully. Physical spaces must accommodate different work modes—quiet zones for focus, collaborative hubs for teamwork, and social areas for relationship building. 📊 Performance management and career development must be location-agnostic: Objective, evidence-based reviews and audits of promotions and pay prevent bias towards in-office employees. Structured mentorship programs provide consistent coaching and growth opportunities regardless of where employees work, supporting equitable professional development. If this helped, please like and share the video. #HybridWork #Leadership #GenZ #RemoteWork #DistributedTeams OUTLINE: 00:00:00 - The Hybrid Disconnect 00:00:49 - Strategic Costs and Leadership Gaps 00:01:43 - Two-Tier Risks and Evidence 00:02:35 - Unseen Saboteurs (Bias and Inconsistency) 00:03:33 - Practical Leadership Solutions

Abstract: Organizations continue to struggle with return-to-office mandates despite clear evidence that younger workers—particularly Generation Z—consistently prefer hybrid arrangements over fully remote or fully in-office models. This article examines the evidence on generational work preferences, the structural challenges facing distributed teams, and the leadership failures that undermine hybrid work effectiveness. Drawing on organizational behavior research and contemporary practice, we identify proximity bias, inadequate manager training for distributed leadership, and executive-employee policy inconsistencies as key barriers to hybrid work success. Evidence-based interventions include structured anchor-day systems with senior leadership modeling, distributed-team management capability building, activity-based workplace planning, and technology infrastructure that equalizes participation. Organizations that treat hybrid work as a leadership and systems challenge—rather than a generational attitude problem—demonstrate better outcomes in talent retention, performance equity, and team cohesion. The article concludes that sustainable hybrid models require deliberate design choices around presence, purposeful co-location activities, and managerial accountability for inclusive team practices.  

In this HCI Webinar, Dr. Jonathan H. Westover talks with Julia Bersin about the recent report out from the Josh Bersin Company, People Management in the Age of AI: The Rise of the Supermanager. Julia Bersin is currently Associate Director, Research at the Josh Bersin Company - studying people practices and technology that help companies transform work for the future. She has a background in B2B tech with a focus on demand gen & growth. She has experience managing multiple functions and teams and marketing to various industries and roles – including HR, TA, Customer Support & Revenue functions.