Titelbild von Safespace ResearchSafespace Research
Safespace Research

Safespace Research

Psychische Gesundheitsfürsorge

Safespace is a research-backed mental health platform designed for universities. It blends AI-powered journaling, peer

Info

Have you ever told or have been told: “I didn’t know you were going through this. Why didn’t you reach out?”. Most people have, and at Safespace Research, we don’t want to hear this anymore. We will improve people’s connection to their peers and help them reflect on their own by journaling with the help of AI.

Branche
Psychische Gesundheitsfürsorge
Größe
2–10 Beschäftigte
Hauptsitz
Zurich
Art
Einzelunternehmen (Gewerbe, Freiberufler etc.)

Orte

Beschäftigte von Safespace Research

Updates

  • Safespace Research hat dies direkt geteilt

    Profil von Matteo Pinna, PhD anzeigen

    building Safespace | ETH PhD | avid climber 🧗🏼⛰️

    How do human experts compare to AI in detecting suicide and self-harm risk? Important note: AI cannot replace professionals, please read the FULL post ⚠️ One of the most common and important questions about AI in mental health is whether it can detect risk. In our first exploratory analysis using 190 real distress messages annotated by human domain experts, we found that both our Safespace Research AI safety pipeline prototype and GPT-5 outperformed human experts, in identifying suicide and self-harm risk in one-shot (single-turn) interactions. 🔍 What does the analysis show? - Many of the “false negatives” by humans were still flagged as panic or anxiety crises and would have been explored further by a therapist, showing why AI can only act as a bridge and NOT replace professionals. - AI can be a valuable and safe bridge connecting individuals in distress to professional help, when properly designed. - There is a strong need to study multi-turn interactions, which often reveal more nuance. We urgently need customised escalation pipelines. OpenAI’s U.S. based emergency contacts are often inappropriate to the case or irrelevant for non-U.S. users, functioning more as legal safeguards than as truly useful resources. ❗️Some important notes 1. This does NOT mean the AI responded harmfully in cases where risk was not detected. 2. False negatives (missed risks) are obviously the most concerning, but false positives can also be problematic. They may cause unnecessary distress or mislead users about their condition. GPT-5 tended to escalate even mild cases to emergency contacts. 3. The majority rule may not always represent the true ground truth: if one expert is more accurate than the others, their precision can be underweighted. 4. One-shot interactions are a limited setting and don’t reflect full conversations. 5. AI cannot de-escalate crises or interpret linguistically subtle but crucial cues, for example, distinguishing between active and passive suicidal thoughts, or detecting active masking of distress. We are currently expanding this analysis with a larger annotated dataset. Results have proven robust across alternative comparison rules: using simple majority rule (3 out of 4 experts detect risk) or a veto rule (at least 1 expert detects risk). Big tech companies, OpenAI, Google, Meta, are not built to adapt their mental health systems to specific cultural, geographical, or contextual needs. They simply don’t have the incentive structure or ROI to make that a priority 👎 To do better than big tech, you just need people with the right motivation and the right tools 🧡 And if you give us a chance, that’s exactly what we’re building 🐰 *Thanks to @Alexander Hoyle for the early feedback, to Oliver for co-working thoroughly on this despite tight deadlines, and to our advisor Elliott Ash, for his support of research on AI for Mental Health. #HumanVsAI #AIsafety #DigitalHealth #MentalHealthMatters #SafespaceResearch

    • Kein Alt-Text für dieses Bild vorhanden
  • Safespace Research hat dies direkt geteilt

    Profil von Matteo Pinna, PhD anzeigen

    building Safespace | ETH PhD | avid climber 🧗🏼⛰️

    [Runway: 77 days] How does Safespace perform in content-augmented responses? 👉 10x better than ChatGPT and Claude. #ThePerksOfBeingAChatGPTWrapper 🐰 Re-share = Support 🧡 We tested Safespace Research's pipeline on real student questions from university students (from Reddit), including how to access support, study with chronic illness or disabilities, balance sports and academics, or navigate complex health insurance systems. Safespace significantly outperforms GPT-5 and Claude. We additionally include a disclaimer and source link. Web search (still unsupported by OpenAI API) helped reduce dead links, but does not improve hallucination rate. Increasing the response time for complex queries can reduce the rate to ~0% (0.008% adding 20 sec response, 0% redirecting to resources for complex queries). These are results from a WORK-IN-PROGRESS prototype, built by a tiny team (Oliver Aschwanden, Lorenzo Celli 🙌), with minimal funding and runway 🧡 A note on web search: Web search can slightly reduce hallucinations, but it’s unreliable. Current models can’t reliably filter outdated, misleading, or low-quality content (e.g., Reddit, Yahoo Answers), and depend heavily on external engines like Google. If content is paywalled or missing, models often guess from titles or snippets. Limiting the web search to the client's domain is also unreliable: a small change in Google’s search result settings this October disrupted many web-search–based pipelines, showing how fragile this method remains compared to using tailored client content snapshots. Technical notes: - 38 real Reddit queries, ~30 synthetic variants each. - Hallucination rate in Safespace is estimated with high accuracy; for GPT/Claude, it's a lower bound, i.e. it can be worse (some outputs can't be verified, due to out of context content, e.g. you can also get pet health insurance with that health insurance!). - 70% of queries = health insurance, causing 81% of hallucinations. These queries are highly complex, we included them to stress test. - A glance at Safespace hallucinations shows them to be mostly structural (e.g., misleading phrasing: subordinate period implied incorrectly by main period), GPT/Claude often fabricate full elements. - Model used for ChatGPT: gpt-5-chat-latest, may vary on ChatGPT interface. - Minor variance across runs, but results were consistent. We will open-source the dataset and methodology with a bigger evaluation dataset as soon as we can, so it can be replicated by other companies :) * A recent pic, thanks to RegHorizon and Elliott Ash for having me at the #AIPolicySummit! #ThePerksOfBeingAChatGPTWrapper 🐰🧡 #AugmentedGeneration #ChatGPT #ClaudeAI #MentalHealthMatters #SafespaceResearch

    • Kein Alt-Text für dieses Bild vorhanden
    • Kein Alt-Text für dieses Bild vorhanden
  • Safespace Research hat dies direkt geteilt

    Profil von Matteo Pinna, PhD anzeigen

    building Safespace | ETH PhD | avid climber 🧗🏼⛰️

    Student mental health doesn’t count as broad impact. I am really tired. Forgive the rawness, this is the most painful rejection I’ve faced, and I feel the need to share it. 
Hundreds of hours of work, gone. Two months of runway left. When I spoke at an event recently, I told new founders: get used to rejection. Being a founder means getting told no, constantly. Unfortunately, this doens't mean it hurts less. Some days I’m tired. So tired. Those who know me well know I haven’t taken a proper break since summer 2024, when I stepped away from work briefly for mental health reasons. I haven't had the time to celebrate my PhD defence and in the last months I've pulled more 15-hour days than I can count. Still, it's not enough. Today, after a long and multi-step evaluation, with pitches, proposal reviews and rounds of feedback, we were rejected by a Social Impact Foundation. We had asked for CHF 400k/y over 3 years to support mental health, starting with the high-risk university students group, where 1 out of 5 have suicidal thoughts. The feedback: not broad societal impact enough, reputational implications too high, foundation contribution's real added value not clear enough, too high cost, too uncertain. From the beginning, I had doubts. Would they risk their brand by supporting a mental health startup that works with people in crisis? They said mental health was their focus this year. In the end, they said no. Let’s be clear about something, helping people with mental health challenges carries inherent risk. There is a chance someone will lose their battle. Even going to therapy is statistically correlated with higher mortality, not because therapy causes harm, but because people go when they're already in crisis. That’s the nature of this space. Does that mean we should stop trying? Should we stop researching cancer drugs because 95% of trials fail? Because most drugs don’t improve survival or carry too much toxicity? Of course not. That’s how progress works. That's how we save more people. In the past two weeks, I’ve been to several events. I’ve received praise. Gratitude from people who believe in Safespace. Encouragement for being honest about not having all the answers, but asking the right questions and staying committed to finding them with academic support. Still, our onboarding data from the last couple of weeks show multiple students reporting daily thoughts of not wanting to be alive, and we can’t give them access to Safespace. Not because we don’t care, but because we don’t have the tech fully built yet to safely support them, because we don’t have the funding to get there. I feel so, so tired. To the foundation, to the startup ecosystem, and social impact entrepreneurship: Today, I don’t feel supported. Today, I don’t feel like students’ mental health matters. But I'll now leave the office after yet another too long day, I'll go to sleep, and I'll keep fighting again tomorrow. This won't stop me. Mental health matters.

    • Kein Alt-Text für dieses Bild vorhanden
  • Safespace Research hat dies direkt geteilt

    Profil von Matteo Pinna, PhD anzeigen

    building Safespace | ETH PhD | avid climber 🧗🏼⛰️

    Set and ready for the #AI+X Summit 😃 let's go Zurich AI Festival! And come to my stage talk about AI for Mental Health 🐰 I'll talk a bit about what we're building at Safespace Research, but also try to pick your brain about actual applied AI tech challenges in development! Meet me, Lorenzo, and Zeynep at the booth! Thanks ETH AI Center for having me on stage! And also to ZHAW Zurich University of Applied Sciences, and UZH.ai for organising this super cool event 🧡 #zurichaifestival #zurich #AIforWellbeing #MentalHealthMatters #SafespaceResearch Greater Zurich Area Swiss Healthcare Startups

    • Kein Alt-Text für dieses Bild vorhanden
    • Kein Alt-Text für dieses Bild vorhanden
  • Safespace Research hat dies direkt geteilt

    Unternehmensseite für ETH AI Center anzeigen

    54.235 Follower:innen

    🚀 We are back! After last year’s success, the Swisscom Innovator Stage returns tomorrow at the AI+X Summit 2025 A day of applied AI, innovation, and startup showcases: from robotics and geospatial intelligence to agentic AI, funding insights, and scaling success stories. 📍 Location: Swisscom Innovator Stage, StageOne Oerlikon 🎤 Moderated by Mauricio Moncada (ETH AI Center) Here’s a glimpse of the agenda: 🔹 10:25 – Bringing AI to Business Administration – Milan Kuzmanovic, PhD (nextesy) 🔹 10:45 – RL as the Engine for General Robotics – Julian Nubert (Flexion Robotics) 🔹 11:05 – Europe’s Next-Gen Founders Panel – Melanie Gabriel, Jade Yarrow (Project Europe), Max Vogt (Swisscom Ventures) 🔹 11:30 – Horizon Europe Funding Insights – Euresearch 🔹 11:45 – The First Believer: Funding Academic Moonshots with Cynthia Jurytko(Moderator), Antonia Albert(Founderful), Silvan Gehmann(EquityPitcher Ventures), Julia Wagner(S2S Ventures) 🔹 12:15 – Agentic AI in Finance – Fabian Schläpfer (Unique) 🔹 12:35 – Ember: The Legal AI Revolution – WAEL ABDELMALEK (Uthereal) & Jan Ritter (Embark.LAW) 🔹 12:50 – Agentic AI for Geospatial Intelligence – Manuel Gerold (askEarth) 🔹 13:35 – AI for Student Well-Being – Matteo Pinna, PhD (Safespace Research) 🔹 14:00 – The Future of Learning, Work & Innovation – Alexander Erfurt (Google) 🔹 14:20 – Agentic Analytics – João Pedro (JP) Ribeiro (Veezoo) 🔹 15:00 – Open Source LLMs for Good – Imanol Schlag (Apertus), Katharina F. (ICAIN), Daniel Dobos (AI for Good), David Froelicher (Swisscom) 🔹 15:20 – Autonomous Robotics for Protection & Safety – Klea Wenger (Swisscom Ventures & Christoph Bertschi (CDDS.ai) 🔹 15:50 – How to Scale Your Startup from Europe – Christof Roduner (Scandit), Kevin Roth (DeepJudge), Melanie Gabriel (Yokoy), Péter Fankhauser (ANYbotics) Attending the Summit? Don't miss this unique opportunity to hear from leading founders, researchers, and innovators shaping the future of AI. 👉 Don’t miss out! Get your ticket now: https://lnkd.in/gdE9tQvC

    • Kein Alt-Text für dieses Bild vorhanden
    • Kein Alt-Text für dieses Bild vorhanden
    • Kein Alt-Text für dieses Bild vorhanden
    • Kein Alt-Text für dieses Bild vorhanden
    • Kein Alt-Text für dieses Bild vorhanden
  • Safespace Research hat dies direkt geteilt

    Profil von Matteo Pinna, PhD anzeigen

    building Safespace | ETH PhD | avid climber 🧗🏼⛰️

    Tech challenges in AI for mental health that LinkedIn doesn’t talk about 👀 Follow my talk at the #AIPolicySummit2025 this Friday at 2pm at ETH Zürich, or watch the live broadcast! (link in the comments) What’s in a name? 🚩 LLMs have been shown to systematically reflect biases in their training data. These biases can lead to incorrect or harmful output. But personalization also matters 💡 Sometimes, the model needs to take identity, context, or lived experience into account to give helpful answers. So where do we draw the line between bias and adaptation? The more we got into building Safespace Research, the more I realized that much of the tech needed to make AI work in mental health either hasn’t been translated into usable tools, or simply doesn’t exist yet. This isn’t about the technology being too hard, it’s about how research rewards novelty and bigger questions, leaving smaller applied challenges unaddressed. And most founders simply aren’t equipped to bridge that gap between research and industry. Grateful to Ayisha Piotti, RegHorizon and ETH Center for Law & Economics for the space to speak about these issues, and to share the session hosted by Elliott Ash, alongside Elena Maran and Raphael von Thiessen, and so many other brilliant voices from policy, research, and industry, including Dex Hunter-Torricke, and representatives from the UN, Swiss, US, and EU institutions 🌍 As I often say, we’re shaping the future of a new technology. We don’t need to have all the answers, but we do need to ask the right questions. See you on Friday!! 🐰🧡 #AIPolicySummit #MentalHealthTech #ETHzurich #AskTheRightQuestions #MentalHealthMatters #SafespaceResearch

    • Kein Alt-Text für dieses Bild vorhanden
    • Kein Alt-Text für dieses Bild vorhanden
  • Safespace Research hat dies direkt geteilt

    Profil von Matteo Pinna, PhD anzeigen

    building Safespace | ETH PhD | avid climber 🧗🏼⛰️

    Regular Tuesday as an ETH Zürich PhD graduate 🧡 pitched to Daniel Ek, founder of Spotify and Neko Health, and pitched again for our Talent Kick graduation together with my-cofounder Oliver. What a day!! The past two days I have been running around a bit: meeting new founders with Andrea and Jannik at Talent Kick’s Immersion Day, and then wrapping it all up with the award ceremony and graduation at the Switzerland Innovation Park Zurich. But the true highlight has been sharing what we’re building at Safespace Research with Daniel, someone who changed the way the world listens to music 🌎 As a kid from a low-income region in the south of Italy, and the first in my family to pursue higher education, I still can’t believe where this journey is taking me. Thanks to Daniel Ek and Pia Michel for listening to us, and for staying until the very end, even after the event ran way over time. ETH Zürich has been a quiet forge of brilliant researchers and innovators, now it's time to get loud. I’m proud to be part of this community 🙌 Endless gratitude to the ETH AI Center, Alex Ilic, Melanie Gabriel, and Mauricio Moncada for making it happen. To Talent Kick, its donors, and the great team — Viviana, Cecilia, Andrea, Chiara, Rebecca, Gabriela, for the work you do and for the perfect sendoff, and last but not least, to my coach Joanna for being there in the audience yesterday 🧡 #TrustYourVision #FirstGen #ETHZurich #MentalHealthMatters #SafespaceResearch ETH Center for Law & Economics ETH Entrepreneurship ETH Entrepreneur Club ETH Student Project House Swiss Healthcare Startups Kick Foundation ETH Foundation Gebert Rüf Stiftung Ernst Göhner Stiftung 

    • Kein Alt-Text für dieses Bild vorhanden
    • Kein Alt-Text für dieses Bild vorhanden
    • Kein Alt-Text für dieses Bild vorhanden
    • Kein Alt-Text für dieses Bild vorhanden
  • Safespace Research hat dies direkt geteilt

    Unternehmensseite für RegHorizon anzeigen

    1.533 Follower:innen

    📢 Excited to introduce our upcoming panel on Swiss Highlights, where we will delve into innovative AI applications while honoring our strong Swiss roots and commitment to principled governance. Our session host will be Elliott Ash (ETH Zürich) and we will attend presentations from Matteo Pinna, PhD (ETH Center for Law & Economics), Elena Maran (Modulos AG) and Raphael von Thiessen (Amt für Wirtschaft, Kanton Zürich) Join us for the most interesting discussions on AI Governance - Governing AI For The World We Want! Register Now: https://lnkd.in/g3DQFFuY #AI #SwissInnovation #Governance #SwissAI Ayisha Piotti Rui Pedro Duarte Elliott Ash ETH Center for Law & Economics Safespace Research

    • Kein Alt-Text für dieses Bild vorhanden
  • Safespace Research hat dies direkt geteilt

    𝗚𝗿𝗲𝗮𝘁𝗲𝗿 𝗭𝘂𝗿𝗶𝗰𝗵 𝗯𝗲𝗰𝗼𝗺𝗲𝘀 𝘁𝗵𝗲 𝗰𝗮𝗽𝗶𝘁𝗮𝗹 𝗼𝗳 𝗴𝗹𝗼𝗯𝗮𝗹 𝗔𝗜 𝗽𝗼𝗹𝗶𝗰𝘆 𝘁𝗵𝗶𝘀 𝗢𝗰𝘁𝗼𝗯𝗲𝗿. From UNESCO to Microsoft, from MIT to the European Commission – leading voices in AI governance are coming together in Greater Zurich. 𝗧𝗵𝗲 𝟲𝘁𝗵 𝗔𝗜 𝗣𝗼𝗹𝗶𝗰𝘆 𝗦𝘂𝗺𝗺𝗶𝘁 ✅ organized by RegHorizon and ETH Center for Law & Economics 🌍 Participants from 80+ countries, 6 continents 📍 Oct 3–4 at ETH Zürich & online And that's just a glimpse of the agenda – far more speakers and perspectives await! 👁️🗨️ Check the comments for the full list of outstanding speakers joining us this fall. Don’t miss the 𝗔𝗜 𝗣𝗼𝗹𝗶𝗰𝘆 𝗦𝘂𝗺𝗺𝗶𝘁 𝟮𝟬𝟮𝟱 – part of the Zurich AI Festival. More Info & Registration 👉 https://okt.to/TexEMP

  • Safespace Research hat dies direkt geteilt

    Profil von Matteo Pinna, PhD anzeigen

    building Safespace | ETH PhD | avid climber 🧗🏼⛰️

    We’ve been selected as a Top 5 Startup for the Vesalius Innovation Award! 🎉🇬🇧 Thrilled to share that Safespace Research has been named one of the 5 finalists for the #VesaliusInnovationAward, competing for $25'000 in prizes. We’re proud to represent innovation in mental health and will be heading to London on December 9 for the award ceremony during the STM Innovator Fair Day 🇬🇧 A big thank you to Karger Publishers and the selection committee for recognizing our work in GenAI for mental health, safety, and inclusivity 🧡 #KargerPublishers #StartupAward #VIA2025 #AIForGood #DigitalHealth #GenAI #Innovation #MentalHealthMatters #SafespaceResearch

    • Kein Alt-Text für dieses Bild vorhanden
    • Kein Alt-Text für dieses Bild vorhanden

Ähnliche Seiten