Matthias Michel
@matthiasmichel.bsky.social
Assistant professor at MIT, Department of Linguistics and Philosophy. Philosophy of science and cognitive science of consciousness.
The result is what I'd predict actually. AI gives high confidence overall because they have global confidence that they can do the task. But they can't track trial-by-trial accuracy. It's metacognitive sensitivity that matters, not overall confidence (wires.onlinelibrary.wiley.com/doi/full/10....).
September 4, 2025 at 11:45 AM
The result is what I'd predict actually. AI gives high confidence overall because they have global confidence that they can do the task. But they can't track trial-by-trial accuracy. It's metacognitive sensitivity that matters, not overall confidence (wires.onlinelibrary.wiley.com/doi/full/10....).
Are you looking for something like this? That's from Laming's Sensory Analysis, based on data from Leshowitz et al. (1968) for luminance difference-thresholds. Older psychophysics papers / books often cite quite a lot of papers showing this kind of results.
September 3, 2025 at 2:22 PM
Are you looking for something like this? That's from Laming's Sensory Analysis, based on data from Leshowitz et al. (1968) for luminance difference-thresholds. Older psychophysics papers / books often cite quite a lot of papers showing this kind of results.
Apparently they made sure to specify 'the use of a calculator is not allowed'.
April 24, 2025 at 12:36 PM
Apparently they made sure to specify 'the use of a calculator is not allowed'.
Light doesn’t travel well underwater. Following MacIver, this severely restricts the visual horizons of aquatic animals—typically to just a few body lengths for ecologically relevant objects. This restricted sensory horizon imposes fast reactions to visual stimuli like predators and preys.
April 21, 2025 at 3:27 PM
Light doesn’t travel well underwater. Following MacIver, this severely restricts the visual horizons of aquatic animals—typically to just a few body lengths for ecologically relevant objects. This restricted sensory horizon imposes fast reactions to visual stimuli like predators and preys.
To illustrate, suppose you’re driving on a foggy road at night. You can’t see very far at all. This means that you’ll have to react very fast to whatever enters your ‘sensory bubble’, or ‘sensory horizon’. Those reactions will have to be driven by unconscious vision. Conscious vision is too slow.
April 21, 2025 at 3:27 PM
To illustrate, suppose you’re driving on a foggy road at night. You can’t see very far at all. This means that you’ll have to react very fast to whatever enters your ‘sensory bubble’, or ‘sensory horizon’. Those reactions will have to be driven by unconscious vision. Conscious vision is too slow.
To illustrate, start with two extreme examples. Suppose you’re visually conscious of a stimulus just 30 milliseconds (ms) after it appears on the screen. This means that consciousness could be involved in very fast (reflex) reactions to visual stimuli.
April 21, 2025 at 3:27 PM
To illustrate, start with two extreme examples. Suppose you’re visually conscious of a stimulus just 30 milliseconds (ms) after it appears on the screen. This means that consciousness could be involved in very fast (reflex) reactions to visual stimuli.
This is the part where the most obvious disanalogy is brushed away as 'beside the point'. Empirical support can't be 'beside the point' when judging whether a theory is scientific or not.
December 11, 2024 at 1:41 PM
This is the part where the most obvious disanalogy is brushed away as 'beside the point'. Empirical support can't be 'beside the point' when judging whether a theory is scientific or not.
In this new preprint @smfleming.bsky.social and I present a new theory of the functions and evolution of conscious vision. This is a big project: osf.io/preprints/ps.... We'd love to get your comments!
November 22, 2024 at 7:20 PM
In this new preprint @smfleming.bsky.social and I present a new theory of the functions and evolution of conscious vision. This is a big project: osf.io/preprints/ps.... We'd love to get your comments!
Preprint coming soon. Almost ready. Very happy about this (ambitious) project!
November 20, 2024 at 5:52 PM
Preprint coming soon. Almost ready. Very happy about this (ambitious) project!
I once received actual mail with this. I think it's a cult.
November 17, 2023 at 6:23 PM
I once received actual mail with this. I think it's a cult.
New paper: we're getting close to showing unconscious perception with a non-biased (2IFC) visibility task, using both metacontrast masking and super fast presentations (980 micro(!!)-seconds) using a modern-day tachistoscope (shown here). Preprint here: osf.io/preprints/ps...
October 16, 2023 at 2:34 PM
New paper: we're getting close to showing unconscious perception with a non-biased (2IFC) visibility task, using both metacontrast masking and super fast presentations (980 micro(!!)-seconds) using a modern-day tachistoscope (shown here). Preprint here: osf.io/preprints/ps...