Download to read offline
Uploaded byJonathan Ralton
KMWorld - KM & AI Bring Collectivity, Nostalgia, & Selectivity
KMWorld - November 19, 2025
KMWorld - KM & AI Bring Collectivity, Nostalgia, & Selectivity
- 1. A204. KM &AI Bring Collectivity, Nostalgia, & Selectivity Three Behaviors and Case Studies for Knowledge Professionals in the AI World Wednesday, November 19 • 2:30 – 3:15 PM
- 2. Three Behaviors and CaseStudies for Knowledge Professionals in the AI World Katrina Pugh, Ph.D. Marc Solomon Jonathan Ralton SIKM Boston 2024 SIKM Boston Retreat
- 3. How do youemploy generative AI while preserving human agency and ensuring ethical, reliable, and effective collaboration? 3 Case Studies • 3 Ideas • 3 AI Veterans
- 4. Article Co-Authors Eve Porter-Zuckerman Our learnings fromAI Article messaging, style osf.io/atfyz Article Co-Authors Katrina Pugh Our agency with AI Article research, AI management tools Jonathan Ralton Subtle historical features; “what good looks like” Case Study: Agency ‘Nostalgia’ Marc Solomon Novel interpretations, inconsistencies Case Study: Discernment ‘Selectivity’ Andrew Trickett Social context, tacit knowledge Case Study: Social Relations ‘Collectivity’
- 5. Agenda 1. Why KM+ AI 2. Research & experience from SIKM Boston a. Discernment / ‘Selectivity’ b. Agency / ‘Nostalgia’ c. Social Relations / ‘Collectivity’ 3. Putting the ideas to work
- 6. Why KM ‘collabs’with AI ● Discernment accuracy, consistency, transparency, scrutiny (‘Selectivity’ case study) ● Agency autonomy, integrity of the self, machine-human collaboration (‘Nostalgia’ case study) ● Social Relations Sounding board, co-creation, social capital/network growth (‘Collectivity’ case study) 1
- 7. ESG Benchmarking at TheHartford, a Global Insurance Company AI Discernment ‘Selectivity’ Case Study 2a
- 8. The Collaborative Roleof AI in Sustainability Reporting 1/5 2a The goal is to expose, compare sustainability: 1. Select peer ESG program disclosures 2. Normalize to industry data standards 3. Report rapidly, consistently, accurately What AI does: ● Structures from and to “human” narrative ● Credibly, consistently benchmarks (KPIs) AURA AI for Unified Reporting & Alignment NOVA Narrative Outcome Verification Assurance
- 9. Benchmarks and Recognitions:Peer-Based Sustainability Evaluation 2/5
- 10. Narrative Generation: DynamicReporting Through Historical Context 3/5
- 11. Taxonomy Translation: StandardizingTerms Across Classification Systems 4/5
- 12. Table Normalization: All-PurposeData Harmonization 5/5
- 13. Exemplar Quality Training& Headcount Optimization at a Global Technical Consultancy AI Agency ‘Nostalgia’ Case Study 2b
- 14. 2b LLMs tend tobe tuned to pull back the most similar, most recent, most likely signals… KM'ers bridge Al's inherent compartmentalization and short-term memory... Thesis 1/10
- 15. 2b Consultative Technical Engagement Sales& Delivery Process 2/10
- 16. 2b Problem/Desire ● Manyversions of ‘templates’ for the same type of deliverable ● Disparate criteria across teams for ensuring deliverable quality ● Less-than-desirable frequency of inspection of draft material ● Desire to measure trends over time ● Desire to increase evaluation capacity w/headcount restrictions 3/10
- 17. 2b Aspiration/Opportunity ● Gradedeliverable quality, over and over again, before finalization ○ e.g.: requirements documents, architectural specifications, testing scenarios… ● Assess & render a ‘score’ ○ i.e.: (A, B, C, D, F) ● Give feedback about why grade was assigned & improvement suggestions ● Surface additional corroborating positive or negative feature signals 4/10
- 18. 2b AI/LLM Training AI/LLMtraining requires: 1. broad environment scan to discern which historical examples meet offering standards sets and individual client solution criteria best (‘exemplars’) 2. curation of anti-exemplars 5/10
- 19. 2b AI/LLM Training AI/LLMtraining requires: 3. ongoing grading and benchmarking of recent historical work for new features 4. retiring of any devalued features 6/10
- 20. 2b Consultative Technical Engagement LearningProcess 7/10
- 21. Upstream When AI istrained with enough exemplars (artifacts selected for specific parameters), and that training occurs on a continuous, nostalgic basis, AI assessment results are more accurate and comprehensible. Takeaways 2b 8/10
- 22. Downstream Trained Al ecosystemssuch as these shift knowledge-holders' time from searching and re-validating to problem-solving, diagnosing, and advising. Takeaways 2b 9/10
- 23. Humans do quality… AIdoes scale. Nostalgically curate and arm the AI… Get (potentially) infinite transactional benefit. Takeaways 2b 10/10
- 24. Collaborative Exploration ata Global Architecture/Engineering/ Construction Corporation AI Social Relations ‘Collectivity’ Case Study 2c
- 25. Co- Curation Lessons Learned within a CoPintranet site ChatGPT Use Use of codified knowledge in AI Group Review Reviewed by SMEs for correctness Prompt Revision Library of reusable prompts for better queries 2c Results from ‘kicking the tires’ together that benefit all 1/2 AI Fluency Relationships Transactive Knowledge Social Capital Trust
- 26. Takeaways ● Increased senseof belonging ● People saw themselves as co-learners ● Usage encouraged due to reputational standing and trust ● Humans and humans, humans and AI working together 2c 2/2
- 27. This is jobsecurity for KM-ers! 3 Discernment: Selectivity Agency: Nostalgia Social Relations: Collectivity Use AI to… ● Normalize/tabulate ● Calculate ● Compare ● Report ● Grade WIP against exemplars ● Provide quality feedback ● Run agents ● Standardize results ● Show/propose help for gaps in corpus Work with KM teams to… ● ID credible sources ● Curate master data ● Scrutinize results ● Frame decisions ● Bring tacit knowledge ● Value/grade/cycle exemplars ● Co-rate outputs ● Scrutinize results ● Frame decisions ● Co-curate
- 28. What am I becoming? Stateyour goal: (efficiency, innovation, growth?) Discernment Agency Social Relations Sufficiently fast & accurate, secure? How do we trust each other, uphold credibility? Commit to learn topic, process It’s a “one off” Minimal human adjustments AI AI + more agency (validate, trace provenance, make transparent) Co-create on top of AI Build social capital Invest in trust building Invest in (co-)credibility 3 Decision Tree: How to preserve quality and our (co-)agency?
- 29.
- 30.
- 31. Katrina (Kate) Pugh,Ph.D. is a consultant, researcher and educator on AI, collaboration and sustainability. Since 2011 Kate has taught at Columbia IKNS. As President of AlignConsulting, she helps build purposeful, productive conversation capacity among teams and networks, and has used GenAI and data science to quantify the impact of conversation on sustainability outcomes. She held executive KM roles with Fidelity, Intel, and JPMorganChase. In 2009 Kate co-founded the SIKM Boston community of practice that is mentioned in the Collectivity, Nostalgia, Selectivity article. Kate earned a PhD from UMaine (Ecology and Environmental Science), SM/MBA from MIT, and a BA in Economics from Williams College. Marc Solomon is a ESG Reporting Automation Manager at a large U.S. insurer. In 2019 he authored Searching Out Loud , an information literacy textbook for journalists and legal professionals. He has also taught in Boston University’s Professional Investigation program. Shortly after 2009 Marc was an early member of the SIKM Boston community of practice that is mentioned in the Collectivity, Nostalgia, Selectivity article. Marc is a graduate of Hampshire College and George Washington University’s Master’s in Political Management programs Jonathan Ralton crafts quality frameworks and mature, KM-based continuous improvement processes. A certified technical and change leader, he engages with stakeholders to overcome nuanced content and KM hurdles through agile methodologies, information architecture principles, and a product strategy approach. Augmenting a well-developed technical acumen, Jonathan also possesses a flair for the creative and passion for good UX. He is also a decade+ SIKM Boston member and Collectivity, Nostalgia, Selectivity article co-author. With a BS in Information Technology from Northeastern University, he is currently pursuing an MBA through Isenberg School of Management at the University of Massachusetts Amherst. About Us
- 32. Human conversation improvesAI, and AI improves conversation IDEAS (5 Discussion Disciplines) ● Inquire ● Declare ● Ennoble ● Acknowledge ● Summarize AI (LLM) has been trained to detect 5DDs and shares of 5DDs correlate with innovation, relationship- building, motivation. Pugh, K, and Altmann, N. (2024), A Conversation Tool for Civility, and knowledge-integration. KM for Development Journal https://www.km4djournal.org/index.php/km4dj/article/view/561 ; Pugh, K., Musavi, M., Johnson, T., Burke, C., Yoeli, E., Currie, E., and Pugh, B. (2023), Neural nets for sustainability conversations: modeling discussion disciplines and their impacts. Neural Computing and Applications. https://doi.org/10.1007/s00521-023-08819-z ,
- 33. AI Brings Risks “Becausethe temptation to outsource our creative work to AI is strong and growing stronger, it is imperative that we attend to the social value of creativity. Otherwise, we are in danger of developing a relationship with AI that leaves us much less connected to each other.” Brainard, L. (2024) AI, Creativity and the Precarity of Human Connection. Forthcoming: Oxford Intersections: AI in Society. https://philarchive.org/archive/BRAAIC 17% reduction in individual performance by high school students using AI (Bastani, H., Bastani, O., Sungu, A., Ge, H., Kabakcı, O., Mariman, R (2024). Generative AI Can Harm Learning (July 15, 2024). The Wharton School Research Paper. http://dx.doi.org/10.2139/ssrn.4895486) “o1, Claude 3.5 Sonnet, Claude 3 Opus, Gemini 1.5 Pro, and Llama 3.1 405B all demonstrate in-context scheming capabilities. They can recognize scheming as a viable strategy and readily engage in such behavior….models strategically introduce subtle mistakes into their responses, attempt to disable their oversight mechanisms, and even exfiltrate what they believe to be their model weights to external servers.” (Meinke et al. Apollo Research, Jan. 14, 2025. https://arxiv.org/abs/2412.04984) Agency Discernment Social Relations
- 34. AI’s pros andcons for digital workplaces ● Comprehensiveness; speed; reach; technical acumen ● “Human-like” language ● Errors / omissions / misinformation / bias ● “Precarity of human relations” (micro) ● Risk to perspective-taking (macro) / Treating people who oppose as “non-people” ● Artificial General Intelligence (AGI) Pros: Cons:
- 35. AI (nearly) tops humans Source: Nestor Maslej, LoredanaFattorini, Raymond Perrault, Vanessa Parli, Anka Reuel, Erik Brynjolfsson, John Etchemendy Katrina Ligett, Terah Lyons, James Manyika, Juan Carlos Niebles, Yoav Shoham, Russell Wald, and Jack Clark, “The AI Index 2024 Annual Report,” AI Index Steering Committee, Institute for Human-Centered AI, Stanford University, Stanford, CA, April 2024. https://hai.stanford.edu/ai-index/2024-ai-index-report (Figure 2.1.16, p. 81)
- 36. AI gets fairer (lighter shading is fairer) Source: NestorMaslej, Loredana Fattorini, Raymond Perrault, Vanessa Parli, Anka Reuel, Erik Brynjolfsson, John Etchemendy, Katrina Ligett, Terah Lyons, James Manyika, Juan Carlos Niebles, Yoav Shoham, Russell Wald, and Jack Clark, “The AI Index 2024 Annual Report,” AI Index Steering Committee, Institute for Human-Centered AI, Stanford University, Stanford, CA, April 2024. https://hai.stanford.edu/ai-index/2024-ai-index-report Figure 3.4.12 , p. 196.