Human-AI Collaboration

Me and my colleagues at Google DeepMind and Google Research are sharing our latest work on tropical cyclone prediction, now available through a research tool, Weather Lab: https://lnkd.in/dNtjmiYq Over the past 50 years, tropical cyclones, also known as hurricanes or typhoons, have claimed more than 779,000 lives and caused $1.4 trillion in economic losses [WMO]. For the millions of people living in their path, the accuracy of weather forecasting is the most critical line of defense. In an effort to protect lives and property from this threat, we’ve built a powerful new machine learning (ML)-based ensemble weather model, deployed it operationally on Weather Lab, and partnered with experts from the U.S. National Hurricane Center (NHC) who will assess its live predictions alongside their established forecasting tools. The ensemble mean cyclone track of our new model gains about 1.5 days of position error advantage over ECMWF ENS in tests based on NHC protocols. And surprisingly, our model has a lower average intensity error than NOAA’s high-resolution hurricane model, HAFS-A, in more than 60 of the 74 cyclones evaluated in 2023 and 2024 in the East Pacific and North Atlantic basins. We achieved this by building a new kind of ML weather model, FGN [Ferran Alet Puig et al., 2025], which substantially outperforms GenCast on probabilistic metrics, and specialising it for cyclone tracking by training it on a record of nearly 5,000 tropical cyclones from the past 45 years. Most human forecasters do not trust a weather model until its performance is demonstrated in a real-time setting. That’s why we built Weather Lab, available globally, providing access to live and historical visualisations of tropical cyclone predictions from our new ML weather model, with WeatherNext and ECMWF models shown for comparison. We recently enabled live data downloads in CSV and ATCF format for experts to evaluate. This is a powerful new tool in the toolbox, but no single model is perfect. It will remain key that human forecasters evaluate a wide range of both ML and physics-based predictions when issuing public warnings for cyclone threats. And of course, ML weather models continue to depend on the historical and real-time availability of atmospheric analysis datasets produced by physical modelling centres, and the continued quality and coverage of the Earth’s observing system. Tropical cyclones will likely become more destructive over time [IPCC, 2023]. It is crucial we continue improving our monitoring, prediction, and understanding of these complex beasts of physics. Try Weather Lab: https://lnkd.in/dNtjmiYq  Blog post: https://lnkd.in/dkj8cYan  FGN (Alet et al., 2025): https://lnkd.in/dJhP9Kj2  WMO: https://lnkd.in/dPt94VX5 IPCC, 2023: https://lnkd.in/dj5n-Rqg 

MIT researchers paired 2,310 people into human-human and human-AI teams to create real ads in a collaborative workspace with some fascinating outcomes—tracking 183K messages, 2m copy edits, and over 5m ad impressions. The paper "Collaborating with AI Agents: Field Experiments on Teamwork, Productivity, and Performance" examined many facets of the dynamics of human-AI collaboration on what was most effective. Some of the valuable insights: 🤖 AI changes how teams talk and work together. Human-AI teams sent 45% more messages than human-only teams, with a focus on task execution—suggestions, instructions, and planning—while human teams sent more social and emotional messages. Despite this shift, both team types rated teamwork quality similarly, showing that collaboration can remain strong even when social interaction drops. 🧍➕🤖 One person plus AI can match or beat human teams. Individuals in human-AI teams produced 60% to 73% more ads than individuals in human-human teams, closing the productivity gap that usually favors groups. Despite having only one human per team, human-AI groups created just as many ads overall as two-human teams. 🧠 Human-AI success depends on psychological compatibility. When a conscientious person worked with a conscientious AI, message volume increased by 62%, signaling better engagement. But mismatches had negative effects—for example, extraverted humans working with conscientious AIs saw drops in text, image, and click quality across the board. 📊 AI lets people shift from doing to directing. Participants in human-AI teams made 60% fewer direct text edits compared to those in human-only teams. Instead of rewriting content themselves, they communicated what needed to be done—refocusing effort from manual changes to guiding and refining AI-generated output. 🔄 AI redistributes cognitive workload and changes who does what. With AI handling routine and complex text generation, humans shifted attention from editing to strategic input and idea generation. This redesigns roles within teams, suggesting new ways to organize work where humans steer, and AI constructs. Humans + AI is the future. This research provides more valuable foundations for understanding how to do this well.

𝐖𝐡𝐞𝐧 𝐇𝐮𝐦𝐚𝐧𝐬 𝐋𝐞𝐚𝐫𝐧 𝐭𝐨 𝐂𝐨𝐧𝐧𝐞𝐜𝐭 𝐀𝐠𝐚𝐢𝐧 In a world where technology scales faster than our ability to sit with another human being and truly listen, I have started to sense a quiet fatigue behind many confident leadership narratives. After a series of conversations in Singapore, including meaningful time with LinkedIn Top Voices Mrs Shamane Tan and Mrs Cassandra Nadira Lee, one insight became impossible to ignore: it is not AI that is running ahead of us, it is our humanity that risks falling behind. This reflection is not about resisting technology. It is about refusing to lead with empty dashboards while trust, empathy, and depth quietly leak out of the system. This article is an invitation to every decision maker who feels that something essential is missing beneath the metrics. It brings together real encounters, uncomfortable questions, and a different kind of blueprint for the future, where data does the heavy lifting, but humans decide the meaning. If you have ever felt that your organization is growing in size yet shrinking in soul, I hope you will read this slowly. The next competitive edge is not more speed. It is the courage to rebuild how we see people, how we design connection, and how we lead with a fully awake conscience.

🌻 Designing For Trust and Confidence in AI (Google Doc) (https://smashed.by/trust), a free 1.5h-deep dive into how trust emerges, how to design for autonomy, risk, confidence, guardrails — with all videos, slides and examples in one single place. Share with your friends and colleagues — no strings attached! ♻️ Google Doc (slides, videos, links): https://smashed.by/trust All slides (PDF): https://lnkd.in/dsq2BAJJ Full 1.5h-video recording: https://lnkd.in/d72b66Qa Zoom video backup: https://lnkd.in/dZJzCnZh Key takeaways: 1. Trust doesn’t emerge by default — it must be earned. 2. Trust means strong believing, despite uncertainty. 3. It’s when system is competent, predictable, aligned. 4. It also means transparency about its limitations / capabilities. 5. AI feature retention often plummets due to lack of confidence. 6. Trust isn’t linear: takes time to be built, drops rapidly in failures. 7. Most products don’t want users to fully rely on them → complacency. 8. Trust requires Understanding + Success moments + Habit-Building. 9. It thrives at intersection of Perceived value + Low cognitive effort. 10. We need to “calibrate” trust to avoid over-reliance and aversion. 11. Transparency only builds trust if users can verify the output. 12. User must feel in control: to validate, shape and override output. 13. Users have low tolerance for mistakes if AI acts on their behalf. 14. High-autonomy + High-risk → human intervention is non-negotiable. 15. Start with human oversight, increase autonomy as trust grows. 16. Perceived usefulness + ease of use are primary drivers of AI adoption.  17. Biggest risk to effort is a blank page → leads to open-intent paralysis. 18. Confidence builds through frequent use, not through “blind” trust. 19. Confidence scores are insufficient to help people make a decision. 20. AI might absorb cognition, but humans inherit the responsibility. Design patterns: 1. Link to specific fragments, not general sources. 2. Show the distribution of opinions, not a final answer. 3. Use structured presets to help articulate complex intents. 4. Rely on buttons/filters for a precise control or tweaking. 5. Show sandbox previews to help understand outcomes. 6. For high-stakes scenarios, design approval steps and flows. 7. Explicitly label the assumptions made during processing. 8. Replace confidence scores with actions, requests for review. 9. Embed AI features into existing workflows where work happens. 10. Proactively ask for context around the task a user wants to do. 11. Reduce effort for articulation with prompt builders/tasks. Recorded by yours truly with the wonderful UX community last week. And a huge *thank you* to everybody sharing their work and their findings and insights for all of us to use. 🙏🏼 🙏🏾 🙏🏾 ↓

This week, California became the first jurisdiction in the world to pass legislation regulating AI companions. Governor Gavin Newsom’s new law, SB 243, sets a precedent that many are calling the beginning of the 'age of AI intimacy regulation.' Under this new law, companies operating in California must introduce safety and transparency measures for AI companions like ChatGPT, Replika, Character.AI, and Meta AI. These include: Clear disclosure that users are speaking to an AI, not a human Safeguards against self-harm or suicidal ideation Special protections for minors, including reminders every three hours that the user is interacting with AI Publicly available crisis-response protocols It’s an extraordinary step because it acknowledges that AI companionship isn’t simply a technical innovation, it’s a social one.  As someone working at the intersection of anthropology, AI ethics, and culture, I see this as a pivotal moment. California has recognised that how we relate to machines can have as much impact on wellbeing as what those machines can do. The question is: can Ireland follow suit? Ireland is potentially becoming one of the world’s most concentrated AI ecosystems, home to global tech headquarters, emerging generative AI startups, and a strong regulatory presence within the EU framework. Yet, we have no clear policy for AI companions, even as platforms offering emotional, romantic, or therapeutic relationships proliferate online. COLONII We’re also uniquely placed to lead. As a nation, we’ve spent decades navigating the intersection between faith, identity, emotion, and social change. We truly understand the cultural weight of intimacy, trust, and moral consequence. We’re also part of the EU’s AI Act, which sets a legal baseline for “high-risk” AI, but that act doesn’t yet deal directly with relational or affective AI, the kind designed to engage our hearts rather than our data. An Irish approach might: Embed AI literacy and emotional awareness into digital education from early years through adulthood (see our campaign: https://ai4i.ai/) Matt Shanahan Angelika Sharygina Associate Professor Ray O'Sullivan et al.) Require emotional safety design audits for AI systems claiming to offer companionship, therapy, or mental-health support. Involve citizens’ assemblies in shaping ethical boundaries for relational AI, because cultural consent matters. Mandate transparency standards that clearly differentiate between an AI that simulates empathy and a human who actually feels it. In Irish research we are currently exploring how these technologies intersect with loneliness, ageing, and mental health, all growing social concerns in Ireland (See Rare TV RTE 'Futureville 2' November 11,12,13 on RTE1) The question isn’t whether AI companions will exist, they already do. The question is how we, as a society, choose to coexist with them. Photo: Tanya Crosbie

This paper analyzes 106 studies to figure out when humans and AI work best together—and when they don’t. 1️⃣ AI + humans isn’t always a winning combo. In most cases, AI-human teams performed worse than the best human or AI alone (average performance drop: -0.23). 2️⃣ AI helps more in creative tasks like writing or designing but hurts in decision-making tasks like choosing between options. 3️⃣ When humans are better than AI, teaming up improves performance. When AI is better than humans, adding humans makes things worse. 4️⃣ Even though AI-human teams don’t always outperform the best individual, they do help humans improve (boosting human performance by 0.64 on average). 5️⃣ AI alone was the most accurate at detecting fake hotel reviews, with an accuracy of 73%. AI + humans together actually did worse at 69%, and humans alone were the least accurate at 55%. This shows that if AI is already better at a task, adding humans can drag it down—possibly because people don’t always know when to trust AI. 6️⃣ Trust is a major issue. People either rely too much on AI (taking its answers at face value) or too little (ignoring good AI advice). 7️⃣ Surprisingly, AI explanations and confidence scores (e.g., “I’m 90% sure”) didn’t help much, even though they’re widely used. 8️⃣ The key to better AI collaboration is designing smarter ways to divide tasks, letting AI handle what it’s best at and humans focus on what they do better. ✍🏻 Michelle Vaccaro, Abdullah Almaatouq, Thomas Malone. When combinations of humans and AI are useful: A systematic review and meta-analysis. Nature Human Behaviour. 2024. DOI: 10.1038/s41562-024-02024-1

AI replaced a large part of coding. But it did not replace engineering. Generating code is becoming easier. Deciding what to build, how systems should work, where risks may appear, and what happens after deployment still requires human judgment. Here are 7 skills developers need to master: → 𝗔𝗿𝗰𝗵𝗶𝘁𝗲𝗰𝘁𝘂𝗿𝗮𝗹 𝗥𝗲𝗮𝘀𝗼𝗻𝗶𝗻𝗴 Turn requirements and constraints into scalable, reliable designs while balancing cost, speed, and trade-offs. → 𝗦𝘆𝘀𝘁𝗲𝗺 𝗗𝗲𝘀𝗶𝗴𝗻 𝗧𝗵𝗶𝗻𝗸𝗶𝗻𝗴 Understand how APIs, services, databases, caches, queues, and monitoring work together in production. → 𝗖𝗼𝗱𝗲 𝗥𝗲𝘃𝗶𝗲𝘄 & 𝗗𝗲𝗯𝘂𝗴𝗴𝗶𝗻𝗴 Check AI-generated code for wrong assumptions, missing conditions, edge cases, performance issues, and production risks. → 𝗣𝗿𝗼𝗱𝘂𝗰𝘁 & 𝗕𝘂𝘀𝗶𝗻𝗲𝘀𝘀 𝗨𝗻𝗱𝗲𝗿𝘀𝘁𝗮𝗻𝗱𝗶𝗻𝗴 Connect user problems and business goals to feature scope, technical decisions, and measurable outcomes. → 𝗦𝗲𝗰𝘂𝗿𝗶𝘁𝘆 𝗔𝘄𝗮𝗿𝗲𝗻𝗲𝘀𝘀 Review permissions, data protection, prompt injection risks, insecure coding, and hidden vulnerabilities before release. → 𝗔𝗜 𝗖𝗼𝗹𝗹𝗮𝗯𝗼𝗿𝗮𝘁𝗶𝗼𝗻 𝗦𝗸𝗶𝗹𝗹𝘀 Give clear context, define tasks well, review outputs, refine prompts, validate results, and reuse proven patterns. → 𝗣𝗿𝗼𝗱𝘂𝗰𝘁𝗶𝗼𝗻 𝗢𝘄𝗻𝗲𝗿𝘀𝗵𝗶𝗽 Monitor logs and metrics, detect incidents, fix root causes, document learnings, and continuously improve the system. AI can generate code quickly. Engineers are still responsible for judgment, reliability, security, architecture, and long-term maintenance. The future belongs to developers who can direct AI and own the outcome. Save this if you are preparing for the next era of software engineering.

The TRIO | Episode 2 | Retaining the Human in Human Resources | Can HR stay human in a world run by data? | In this digital-first, AI-powered world, organizations are innovating faster than ever—but here’s the question we’re diving into: Is HR evolving… or evaporating into dashboards and algorithms? This episode brings together three Human Capital thought leaders who believe HR is not just a function—it’s a force. A force to connect, to care, and to lead with purpose over process. 3 Big Questions We’re Tackling: 1)Can HR still be the heart of the organization when tech threatens to take center stage? (Dr. Sujatha Muthanna | Lead Infosys Learning Advisory) 2)How do we use AI and data to serve people—not just processes? (Sanjiv Agarwal | HR Head, Swiss Re, South East Asia) 3)What does it mean to lead with purpose in HR—and not get lost in the tech maze? (Nathan SV | Co-Founder and Chairman, Visara Human Capital Consulting) Plus: A rapid-fire guest-to-guest Q&A, where insights collide, assumptions are challenged, and perspectives evolve. This episode isn’t about resisting change—it’s about leading change while keeping humans at the center of everything HR stands for. Your Turn: As we prep for the drop—what do you think is the #1 risk of going too far down the AI path in HR? Comment below —we’ll pick a few thoughts to feature in the episode recap! #FutureOfHR #HumanAtTheCenter #AIandHR #TheTrio #TheMondayMorningLearningPodcast #SaiAcuity #India

Tired of AI projects that don't deliver? Try this human-centred approach. From my research over the past couple of years, I’ve noticed a recurring pattern. We often treat AI as a technology experiment rather than an upgrade to how people actually work. That mindset can quietly limit a project’s success. To support better decisions, I’ve developed a human-centred AI readiness checklist based on that research. I hope it’s useful for your next initiative. 𝗦𝘁𝗿𝗮𝘁𝗲𝗴𝘆 𝗮𝗻𝗱 𝗢𝘂𝘁𝗰𝗼𝗺𝗲 𝗖𝗵𝗲𝗰𝗸 (𝗖𝗥𝗜𝗦𝗣-𝗗𝗠 𝗺𝗶𝗻𝗱𝘀𝗲𝘁) →Are we clear on the operational outcome and metric we are improving? ↳If we cannot say “this reduces X by Y%”, we are chasing tools, not performance. 𝗗𝗲𝗰𝗶𝘀𝗶𝗼𝗻 𝗠𝗮𝗽𝗽𝗶𝗻𝗴 𝗖𝗵𝗲𝗰𝗸 (𝗟𝗲𝗮𝗻 𝘄𝗼𝗿𝗸𝗳𝗹𝗼𝘄 𝘁𝗵𝗶𝗻𝗸𝗶𝗻𝗴) →Which real human decisions are we supporting? ↳AI should strengthen judgment points like prioritisation or scheduling, not automate activity without purpose. 𝗣𝗿𝗼𝗰𝗲𝘀𝘀 𝗦𝘁𝗮𝗯𝗶𝗹𝗶𝘁𝘆 𝗖𝗵𝗲𝗰𝗸 (𝗟𝗲𝗮𝗻 𝗽𝗿𝗶𝗻𝗰𝗶𝗽𝗹𝗲) → Is the workflow stable enough to augment? ↳Automating instability scales, defects and frustrates the people doing the work. 𝗩𝗮𝗹𝘂𝗲 𝘃𝘀 𝗗𝗶𝘀𝗿𝘂𝗽𝘁𝗶𝗼𝗻 𝗖𝗵𝗲𝗰𝗸 (𝗣𝗼𝗿𝘁𝗳𝗼𝗹𝗶𝗼 𝘁𝗵𝗶𝗻𝗸𝗶𝗻𝗴) →Does the benefit outweigh frontline disruption? ↳Operational AI should improve flow, not create friction for teams. 𝗗𝗮𝘁𝗮 𝗥𝗲𝗮𝗹𝗶𝘁𝘆 𝗖𝗵𝗲𝗰𝗸 (𝗖𝗥𝗜𝗦𝗣-𝗗𝗠 𝗱𝗮𝘁𝗮 𝘂𝗻𝗱𝗲𝗿𝘀𝘁𝗮𝗻𝗱𝗶𝗻𝗴) →Does our data reflect lived operational reality? ↳Human trust collapses when AI runs on distorted inputs. 𝗛𝘂𝗺𝗮𝗻 𝗖𝗼𝗻𝘁𝗿𝗼𝗹 𝗖𝗵𝗲𝗰𝗸 (𝗛𝘂𝗺𝗮𝗻-𝗰𝗲𝗻𝘁𝗲𝗿𝗲𝗱 𝗔𝗜 𝗱𝗲𝘀𝗶𝗴𝗻) →Where does AI advise, where do humans review, and where does automation act? ↳Clear boundaries protect autonomy and accountability. 𝗥𝗶𝘀𝗸 𝗮𝗻𝗱 𝗥𝗲𝘀𝗶𝗹𝗶𝗲𝗻𝗰𝗲 𝗖𝗵𝗲𝗰𝗸 (𝗡𝗜𝗦𝗧 𝗔𝗜 𝗿𝗶𝘀𝗸 𝗺𝗼𝗱𝗲𝗹) →Have we planned for failure, overrides, and fallback workflows? ↳Operations must remain safe and continuous when systems misfire. 𝗢𝘄𝗻𝗲𝗿𝘀𝗵𝗶𝗽 𝗖𝗵𝗲𝗰𝗸 (𝗢𝗽𝗲𝗿𝗮𝘁𝗶𝗻𝗴 𝗺𝗼𝗱𝗲𝗹 𝗰𝗹𝗮𝗿𝗶𝘁𝘆) →Who owns outcomes, model behaviour, and data quality? ↳Human accountability must remain visible after launch. 𝗜𝗻𝘁𝗲𝗴𝗿𝗮𝘁𝗶𝗼𝗻 𝗥𝗲𝗮𝗹𝗶𝘁𝘆 𝗖𝗵𝗲𝗰𝗸 (𝗦𝘆𝘀𝘁𝗲𝗺𝘀 𝘁𝗵𝗶𝗻𝗸𝗶𝗻𝗴) →Will this support how people actually work? ↳Tools that slow teams are quietly abandoned. 𝗔𝗱𝗼𝗽𝘁𝗶𝗼𝗻 𝗮𝗻𝗱 𝗧𝗿𝘂𝘀𝘁 𝗖𝗵𝗲𝗰𝗸 (𝗖𝗵𝗮𝗻𝗴𝗲 𝗱𝗶𝘀𝗰𝗶𝗽𝗹𝗶𝗻𝗲) →Are we designing for understanding, transparency, and behavioural adoption? ↳Trust grows when teams see AI improving their work, not replacing it. AI is an amplifier. It scales what we already have: good or bad ↳𝐆𝐚𝐫𝐛𝐚𝐠𝐞 𝐢𝐧. 𝐀𝐦𝐩𝐥𝐢𝐟𝐢𝐞𝐝 𝐠𝐚𝐫𝐛𝐚𝐠𝐞 𝐨𝐮𝐭.⁣ ⁣⁣⁣⁣⁣⁣⁣⁣⁣⁣⁣⁣⁣ ⁣⁣⁣⁣⁣⁣⁣⁣The strongest AI initiatives aren’t just technology deployments. They are human-centred operating upgrades that happen to use AI. ♻️ Share if you found this useful. #AIinBusiness #HumanCenteredAI #Operations #Leadership #AIStrategy

As a CMO, one of my top priorities right now is working out what role AI will play in our marketing work at Google. In my experience, Creatives are always the first to play with new tools, and AI is the most exciting sandbox yet. I believe this moment could be a fundamental shift in how we create, allowing us to have impossible ideas and to do things we never could before. While it is still early days, we are already seeing AI revolutionise our workflows, whether it's saving us countless hours storyboarding with ImageFX, or generating 300 variations in one day of our Best Phones Forever spots, using AI to generate copy and visuals. AI can help us do creative testing way faster, or respond to a brief with lots of ideas (or help us organise all the ideas we had but never shipped). Building a culture of experimentation on my team has always been a top priority. Now everyone, regardless of their role on the team, can make things and bring their ideas to life. And the most important part is that humans are in control. We are still the ones calling the shots and making sure that the final work we put out into the world meets our high bar. AI just helps us get there faster, bolder, and with more fun toys along the way. Exciting times! I really enjoyed chatting with Fast Company and Jeff Beer about how my team is harnessing #AI across every stage of the creative process, from ideation to creation. Check out our full conversation & let me know how you’re using AI in your creative process: https://lnkd.in/gxvv2UBD