I enjoyed the paper and there are several great points like that trust is a human-evaluated concept. However the population were Mechanical Turkers, a general population. I'd expect experts like statisticians and data scientists to perform very different.
Remarkable finding: people don't trust transparent models any more than opaque ones, and have more difficulty detecting large errors in transparent ones: https://arxiv.org/pdf/1802.07810.pdf …
-
-
-
An expert can engender and embody trust through a number of ways (or lose it!). An expert can be accountable in a way an algorithm never can.
Keskustelun loppu
Uusi keskustelu -
-
-
Very interesting. I am curious about a followup question: Choose best way to fill in the blank: I _____an ML researcher and I find this result ____ surprising
-
(i realize that this survey is not very clearly presented, results will be biased, etc.. doing it right would be a small project... but it was more fun to make the comment as a survey rather than explicitly :)
Keskustelun loppu
Uusi keskustelu -
-
-
There are two ways of constructing a software design: One way is to make it so simple that there are obviously no deficiencies, and the other way is to make it so complicated that there are no obvious deficiencies. The first method is far more difficult. ― C. A. R. Hoare
-
This is a fantastic quote.
Keskustelun loppu
Uusi keskustelu -
-
-
Transparency in models isn’t about “trust” it’s about verifiability and repeatability and also teachability so humans can learn from how the model is coming to its conclusions. To state that people dont trust them more is a strawman argument. And misses all these other points.
-
Tämä twiitti ei ole saatavilla.
- Näytä vastaukset
Uusi keskustelu -
-
-
Kiitos. Käytämme tätä aikajanasi parantamiseen. KumoaKumoa
-
-
-
Any study in
#explainableAI focusing on error instead of confusion makes me go hmm. The task people were asked to do here isn't the kind where explainable AI shines. Let's talk about scenarios. They matter. I use math ed & have no trouble pulling these groups apart. -
There is a ton of work on formative assessment & how effective it is in learning protocols. So why didn't the components of formative assessment get assembled by the subjects here? That's the question. I suspect because scenario selected is a training exercise not a learning one.
- Näytä vastaukset
Uusi keskustelu -
Lataaminen näyttää kestävän hetken.
Twitter saattaa olla ruuhkautunut tai ongelma on muuten hetkellinen. Yritä uudelleen tai käy Twitterin tilasivulla saadaksesi lisätietoja.