Microsoft Research - Celebrating Susan Dumais: Reflections on a Legacy of Research and Collaboration | Plenary Session
The session highlights Sue's journey from her early academic days to her impactful career at Microsoft. Initially interested in law, Sue shifted to psychology, influenced by a course in mathematical psychology. Her work at Bell Labs and later at Microsoft focused on human-computer interaction and information retrieval, contributing significantly to AI and cognitive science. Her research on latent semantic indexing and personalization in search has been foundational. Sue's interdisciplinary approach bridged AI and HCI, fostering innovation and collaboration. Her mentorship, especially for women in tech, is celebrated, alongside her ability to integrate fun and creativity into her work, enhancing productivity and innovation. Her projects, like News Junkie, exemplify her ability to apply rigorous methodologies to real-world problems, making significant impacts on Microsoft's products and services.
Key Points:
- Sue's interdisciplinary approach bridged AI and HCI, fostering innovation.
- Her mentorship has been pivotal, especially for women in tech.
- Her work on latent semantic indexing and personalization in search is foundational.
- Projects like News Junkie show her ability to apply rigorous methodologies to real-world problems.
- Her integration of fun and creativity into work enhances productivity and innovation.
Details:
1. 📸 Welcoming & Introductions
- Encourage participants to sign the photo book to foster engagement and create memorable experiences, ensuring a personal touch as they arrive.
- Highlight the significance of having Sue sign the proceedings to add perceived value and importance, potentially increasing the event's prestige and memorability.
2. 🎥 Plenary Session Kickoff
- The plenary session, an essential component of the event, is recorded and broadcast to reach a wider audience beyond those physically present. This approach ensures inclusivity and accessibility, allowing participants from various locations to engage with the session's content.
- The session is strategically designed to encapsulate key insights and discussions that set the tone for the day's activities. By broadcasting this session, the organizers aim to maximize reach and impact, fostering a sense of community and shared learning among attendees and remote participants.
- This initiative not only broadens the event's accessibility but also enhances the dissemination of critical information and ideas presented during the session.
3. 👩🔬 Sue's Spontaneous Reflections
- Sue's reflections are impromptu, emphasizing her authentic engagement with the audience.
- She shares real-time thoughts and experiences, illustrating the practical application of her work.
- Sue discusses specific examples from her recent projects, providing actionable insights into her methodologies.
- Her session includes metrics from her latest research, such as a 20% increase in efficiency using a new analytical tool.
- Sue highlights the importance of adaptability in research, reflecting on how spontaneous problem-solving has led to innovative outcomes.
4. 🌱 Sue's Academic Journey & Influences
- Sue's academic journey was significantly shaped by key influences and meticulous planning, which were pivotal in developing her research focus.
- She attributes her growth to specific mentors and educational experiences that provided her with the necessary tools and insights.
- Sue's planning involved setting clear academic goals and seeking out opportunities that aligned with her research interests.
- Her journey includes transformative experiences during her undergraduate studies that led her to pursue a career in research.
- Mentorship played a vital role, offering guidance and support that influenced her research direction and personal growth.
5. 📚 Mentorship & Early Career
- Sue's graduate thesis advisor was crucial in shaping her career path, exemplifying the significant role of effective mentorship during early career stages.
- Rich Shiffron, a distinguished psychologist and National Academy member at Indiana University, exemplifies the impact of learning from industry leaders.
- Mentorship from Rich Shiffron not only provided academic guidance but also opened opportunities in the field of psychology, demonstrating the strategic advantage of being mentored by established professionals.
- Sue's personal experiences highlight that mentorship can offer specialized knowledge and networking opportunities, which are pivotal for career advancement.
- Rich Shiffron's mentorship style, characterized by a focus on research excellence and professional development, significantly influenced Sue's career decisions, particularly in research orientation.
6. 🔍 Sue’s Research Themes & Contributions
6.1. Sue’s Research Focus
6.2. Sue’s Influence and Recognition
7. 🚀 Career Milestones at Microsoft
- The speaker expresses deep gratitude for the recognition received in AI contributions.
- Despite unexpected timing, the speaker appreciates the overwhelming support from both in-person attendees and online participants.
- Acknowledgment of the event organizers for their efforts, mentioning their surprise at the event's details being unknown until the last moment.
8. 🔧 Innovations in Information Retrieval
- Innovative environments blend fun, scientific ideas, and technology, fostering a culture of humor and creativity, essential for success in tech fields.
- Career paths often evolve in unexpected ways, as initial aspirations may differ from current roles, exemplified by the transition to a technical fellow at Microsoft and director of East Coast labs.
- The rapid evolution of technology companies and the rise of fields like generative AI present unprecedented opportunities, underscoring the importance of adaptability and openness to new possibilities.
- Reflecting on career journeys can reveal how unexpected opportunities align with personal and professional growth, emphasizing the value of flexibility in career planning.
9. 🧠 Cognitive Science & Technology Integration
- The speaker initially pursued environmental law but shifted focus after excelling in math and discovering a mathematical psychology course.
- This course, taught by Richard Fry, introduced foundational work by Atkinson, Bower, and Kurthers on attention and memory.
- These studies significantly advanced the understanding of long-term and short-term memory frameworks, influencing cognitive science.
- Cognitive science principles from this research are applied in technology today, enhancing machine learning algorithms and AI-driven systems.
10. 🔄 Evolution of Research at Microsoft
- The integration of quantitative methods and computer simulations to model human intellect, memory processes, learning, and vision was a key attraction for potential researchers, significantly impacting the direction of research at Microsoft.
- A significant career shift occurred when a researcher moved from pursuing law to a double major in psychology, driven by an interest in applying technology and mathematics to solve real-world problems, illustrating how personal interests can shape professional trajectories.
- Research conducted on visual illusions and perceptual learning was pivotal, resulting in a published paper in 'Nature', marking its 50th anniversary, highlighting the long-term impact of foundational research.
- A notable anecdote from the research includes a confusion between 'casually' and 'causally' related concepts, serving as a unique reminder of the importance of precision in academic publications.
- Early research experiences, particularly in mathematical psychology, emphasized the significant role that undergraduate research can play in shaping career paths, underscoring the strategic value of early academic exploration.
11. 🔗 Interdisciplinary Approaches & Collaborations
- The speaker highlights their background in cognitive science and their work at Indiana University, which was a hub for research in perceptual learning and automatic vs. controlled processes.
- They transitioned to Bell Labs, co-founding a human-computer interaction group, despite having no prior industrial research experience.
- Their work at Bell Labs coincided with the digitization of information, including the yellow and white pages moving online.
- They contributed significantly to latent semantic indexing (LSI), which addressed the mismatch between authors' language and that of information seekers.
- LSI reduced large word spaces into smaller, semantic-rich variables and influenced early search engine technology and later probabilistic neural networks.
- The work on LSI informed foundational developments in probabilistic models and neural nets, impacting the field of information retrieval and natural language processing.
12. 💡 Human-Centric System Design
- The server named warden.col.edu at Colorado signifies continuity with past projects, illustrating the importance of legacy in system design.
- Early systems utilized 'pico language models' with limited data, unlike today's expansive, rich models, highlighting the evolution in data and computational capabilities.
- Current systems benefit from large-scale data and improved representation richness, allowing for enhanced system capabilities and performance.
- Tom’s 28-year tenure at Microsoft emphasized interdisciplinary collaboration, leading to innovative re-imaginings in decision theory, graphics, and NLP.
- Collaborative work with Eric, David Hecerman, and Sahami on email spam filtering pioneered advanced text classification using support vector machines, showcasing the impact of teamwork in rapid development.
- Within 36 hours, they developed what was then the world's best text classification system, demonstrating the power of collaboration and swift innovation.
- The understanding of 'hapax logina' in computational linguistics informed improved sampling and filtering strategies in text classification, exemplifying the application of linguistic insights in practical system improvements.
13. 📝 Reflections on User Interaction & System Insights
13.1. User Interaction & System Insights
13.2. Log Analysis and User Behavior
14. 🔍 Finding and Refinding Information
- To fully understand user behavior, it's crucial to complement web search logs with methods like deep ethnography, surveys, and small laboratory experiments, providing context beyond mere search actions.
- Large-scale logs offer valuable insights but need to be integrated with other evaluation methods to capture a complete picture of user needs and frustrations.
- Research methodologies and system designs must adapt to the ever-evolving technology landscape and shifting user aspirations.
- Refinding old information is as important as finding new information, a concept emphasized in Jaime's thesis work.
- The development of tools such as the 'curious browser' helps link implicit user behavior with explicit actions, enhancing understanding of user interactions.
- Addressing the challenge of bridging unlabeled and labeled data is critical for effective system building and deployment.
- Early prototypes like 'stuff I've seen' were created to aid users in refinding previously accessed information, highlighting practical applications of these insights.
15. 📈 The Dynamics of Information & Personalization
- Information silos have become more pronounced due to the proliferation of devices and applications, making it essential to address this growing problem, especially with generative models where context is often lost.
- Personalized search has evolved by considering variables like individual user profiles, metadata, and contextual factors, highlighting the importance of understanding the 'who', 'when', and 'what' of search queries.
- Temporal dynamics of information are critical as the world is dynamic, and protocols often treat it as static. Projects like 'news junky' have explored how stories evolve over time, emphasizing the need for understanding information needs as they change.
- The work on 'refinding' information and contextual understanding in search has expanded beyond traditional web search, including areas like email search, showing significant differences and the necessity for tailored methodologies.
- Collaborative efforts and discussions have been pivotal, underscoring the significance of interdisciplinary work and the importance of people and ideas in driving innovation.
16. 🎓 Rich Shiffron’s Reflections on Sue’s Graduate Days
- Rich Shiffron emphasizes the importance of taking time to celebrate successes and enjoy the journey, as reflected in a recent MSR off-site event.
- Rich parallels his own entry into the field of psychology with Sue's journey, highlighting the influence of diverse academic backgrounds.
- Sue is recognized for her mentorship, particularly in empowering young women at Microsoft, helping them succeed in male-dominated fields.
- Sue's role as a role model extends beyond her scientific accomplishments, impacting women in science, business, and society.
17. 🎓 Rich's Continued Reflections & Sue's Dissertation
- Sue graduated with a PhD 46 years ago under Rich's supervision.
- Sue's dissertation research placed limits on learning in non-consistent training contexts using a visual search task.
- The dissertation suggested that learning an action in different contexts can lead to varied levels of mastery, with some actions being well-learned, others moderately, and some not at all.
- Rich admits to not remembering much but acknowledges Sue's work as highly relevant and potentially publishable even today.
- Sue was awarded the Richard C. Atkinson Lifetime Achievement Award from the Psychological and Brain Sciences Department at IU.
- Rich acknowledges the challenges in mentorship, including ethical issues like author listing, but credits Sue for her successful independent career.
- Sue's work at Bell Labs and Microsoft contributed significantly to practical applications in society, diverging from Rich's basic science focus.
- Rich reflects on the importance of independent thinking in academic and professional growth as evidenced by Sue's career.
18. 🎉 Celebrating Sue’s Legacy & Contributions
- Susan played a pivotal role in shaping Microsoft's research direction, focusing on information retrieval and search, which were considered emerging fields at the time.
- She was instrumental in founding the Adaptive Systems and Interaction (ASI) team, which bridged AI and HCI, fostering cross-disciplinary creativity for 20 years.
- Susan's approach was characterized by a rigorous, human-centered perspective, which redefined problem-solving methodologies.
- She helped expand the ASI team into the Clues (Context, Learning, and User Experience for Search) team, nurturing future leaders like Jamie Tivon and Ryan White.
- Susan's teams produced frontier research and prototypes, integrating insights from multiple disciplines to address sociotechnical issues and human-centric design.
- She was a connector and bridge builder, fostering strong relationships with product teams and breaking down silos within Microsoft.
- Susan's mentorship was recognized by the ACM Athena Lecturer award in 2014 for advancing women in computing.
- She has nearly 100,000 citations, with 20 papers having over 1,000 citations each, highlighting her influence in scholarly work.
- Susan received numerous awards, including the SIGIR Gerard Salton Award for lifetime achievement and the SIGAI Lifetime Research Award.
- In addition to her scholarly contributions, Susan is known for creating an environment where humor and discovery went hand in hand.
19. 😊 Sue’s Impact on Creativity and Collaboration
- Sue has a unique ability to infuse joy and laughter into her work, which serves as a catalyst for creativity and innovation in collaborative environments.
- Research in cognitive psychology shows that a positive mood enhances creative problem-solving abilities.
- In one study, positive moods were induced by showing participants humorous movies, leading to better performance on challenging tasks like the Dunker's candle task.
- Sue’s project, News Junkie, explored the temporality of information in the early 2000s, impacting Microsoft's news services.
- The project used mathematical methods like Kullback-Leibler divergence and Jensen-Shannon divergence to personalize news feeds based on novelty and relevance.
- Sue’s approach involved defining clear metrics, performing rigorous evaluations, and maintaining a fun process throughout.
- The News Junkie project provided a personalized news experience by ordering news articles to add maximum information and studying topic evolution over time.
- Sue has contributed significantly to scientific guidance and mentoring, bringing interdisciplinary, user-centered perspectives to various projects.