Federico Cabitza

@cabitzaf

Health Informatics scholar (PhD), Assistant Professor of Human-Computer Interaction & Data Visualization @ University of Milano-Bicocca, Italy.

যোগদান করেছেন মে ২০১৭

মিডিয়া

  1. ১২ ঘন্টা আগে

    Only 10-15% of readers of interactive visualizations on the New York Times actually click buttons. Maybe, as suggested by , "dataviz people spend too much time thinking about the interactions themselves and less about the audience who is supposed to be using them."?

    এই থ্রেডটি দেখান
  2. ১৫ জুলাই
    উত্তর দিচ্ছেনReplying to এবং অন্য

    On JAMA, 7 months ago, we wrote this. Glad to see that the front is widening and strengthening.

  3. ৭ জুলাই

    A few questions by to ponder carefully, before even hoping, let alone advocating, that AI-powered decision support technology will spread in clinical settings any time soon.

  4. ৬ জুলাই
    -কে উত্তর দিচ্ছেন

    In this sense, A "book of Because" is still to be written, and much hoped for...

  5. ১ জুলাই

    I don't mean to be nominalistic, but is a misnomer wrt . Bias doesn't lie in algos: in fact it's all in the data itself. And ultimately it's us who discriminate people, taking the machine output at face value. It's machine-induced bias, or credulity.

  6. ৩০ জুন

    Explainability advocates seem to overestimate the capability of users to make sense of AI explanations. However these are but other data, output metadata,in the absence of a causal explanatory framework. Distrusting a black box is easier than a white box. Glassy automation bias.

    এই থ্রেডটি দেখান
  7. ৩০ জুন

    A conjecture re algo explainability: What if making medical AI more explicable made it generally more convincing, and hence more effective in inducing errors and bias when it's inaccurate? Plausible explanations may be decisive, but also for the worse. The white box paradox.

    এই থ্রেডটি দেখান
  8. ২৬ জুন
  9. ২৬ জুন
    -কে উত্তর দিচ্ছেন

    Or just short-sighted *technophilia*. (cf. "Close to the machine: technophilia and its discontents" by Ellen Ullman).

  10. ২০ জুন
    -কে উত্তর দিচ্ছেন

    Who knows? Good to put question marks to this and similar claims. As practice makes perfect, so *lots* of data work makes AI perfect. Moreover, historically automation's always been aimed at making work processes more efficient, or more effective, or both. Not really more humane.

  11. ১৮ জুন
    -কে উত্তর দিচ্ছেন

    AI is currently considered a certainty dispenser. In this role, it plots against our knowledge even worse than sheer ignorance, like an unaware misinformer. Just a spoonful of prediction intervals and confidence intervals would help people produce AI antibodies and live happily.

  12. ১৭ জুন
    -কে উত্তর দিচ্ছেন

    It made me recall this "setup", which is as accurate as our best AI to date in the task of distinguishing benign from malignant human breast histopathology (source: ). It never stops making me think... 🤔

  13. ১৬ জুন
    -কে উত্তর দিচ্ছেন

    That's the point!As a non-radiologist working in AI I can even claim this is the most basic and essential assumption in the field. AI looks like "a giant with feet of clay": title of a chapter by (), available on arxiv: .

  14. ১২ জুন

    I'd argue that we need to make more clear the distinction btw algorithmic AI and situated AI. Situated means more than applied, it's about AI embedded in practices of work, relation and interpretation of the reality surrounding us. From the myth of accuracy to ethos and praxis.

  15. ৩ জুন
    উত্তর দিচ্ছেনReplying to এবং অন্য

    Star Trek (TOS 2x24), 8 March 1968

  16. ২ জুন
    -কে উত্তর দিচ্ছেন

    This story is nice doc locker-room humour, but I'm afraid it's misleading in its deepest implications. Telescopes don't label what they "see", while AI does just the opposite (w/ exceptions), it gives labels without telling why. Below 2 slides from a talk at .

  17. ২ জুন
    -কে উত্তর দিচ্ছেন

    At the NYU, the day we met, I showed this slide, just to try orient myself in the forest of synonyms. But I know, entia non sunt multiplicanda praeter necessitatem. 🙄

  18. ২ জুন
    -কে উত্তর দিচ্ছেন

    A *vita activa* is a life in which each of us is thrown into the uncertainty of real life, where to take active part in looking for sensible questions. And where no oracle can relieve us of the responsibility of coming up with meaningful answers. Free will, to mistake.

  19. ১ জুন
  20. ৩১ মে
    -কে উত্তর দিচ্ছেন

    Indeed! Humans are quite bad at many things, why making fair and accurate algorithms should be an exception? Just a joke. More seriously: AI is advocated to give us solutions, answers, advice. Don't we see a prospect of adolescent recklessness in it? Passive life vs vita activa.

লোড হতে বেশ কিছুক্ষণ সময় নিচ্ছে।

টুইটার তার ক্ষমতার বাইরে চলে গেছে বা কোনো সাময়িক সমস্যার সম্মুখীন হয়েছে আবার চেষ্টা করুন বা আরও তথ্যের জন্য টুইটারের স্থিতি দেখুন।

    আপনিও পছন্দ করতে পারেন

    ·