Talk:Cognitive science

From RationalWiki
Jump to: navigation, search

Two thousand years of Western thought overturned[edit]

The quote is convoluted & jargon-heavy, & the citation is to a hardcopy book, leaving this section rather opaque. Can this argument be summarised instead in layman's terms? WěǎšěǐǒǐďWeaselly.jpgMethinks it is a Weasel 11:59, 15 January 2013 (UTC)

It boils down quite simply to "disregarding the role of the body (and how its sensorimotor apparatus motivates metaphorical reasoning) leads to philosophical error." TOW has masses of material on the subject; wp:Embodied cognition is as good a place as any to start. Sprocket J Cogswell (talk) 15:00, 15 January 2013 (UTC)

Why the focus on embodiment?[edit]

Not only is embodied cognition far from mainstream cognitive science, but it's also far from empirical science in general. dildzzzz 15:29, 20 April 2014 (UTC)

I'm going to assume that no answer means nobody minds/cares and at some point I will update this to be oriented more towards mainstream cogsci. Femilisk is watching 09:14, 24 May 2014 (UTC)
Not empirical? How so? My limited understanding is that the embodiment stuff comes in by way of linguistics, and can be verified with things like the susceptibility of reaction time to priming. Maybe its old news, idk.
Nebuchadnezzar might be able to shed some light on this. You watching, Neb? Sprocket J Cogswell (talk) 14:04, 24 May 2014 (UTC)
Embodiment is problematic in this article specifically for two reasons: a. it doesn't represent all of mainstream cog(neuro)sci, and b. it really has failed to contribute something new over and above very very vague verbal theorising. It does not contribute new predictions (if any) nor any new accounts; it just repackages them (which you can imagine is problematic too). If you need references I have quotes from Barsalou (one of the main players of embodied cognition) directly admitting some of the shortcomings of this approach. I also have a person in my lab who has tried (and failed) to gather empirical data countless times that support (only) grounded/embodied/situated/etc. I.e., they tried to gather data that do not support others theories also, only the embodied account, but it's impossible due to the vagueness of the account. I also have a section I can copy/paste from my thesis on the numerous problems with these accounts, but not until I submit it since I don't want sections of my thesis to show up on the internet before submission. But I rewrite it, if you really need me to. Femilisk watching.gif Shhnotearsonlydreamsnow 09:42, 25 May 2014 (UTC)
PS: "embodiment stuff comes in by way of linguistics, and can be verified with things like the susceptibility of reaction time to priming." - yes, that's exactly why it's problematic. RTs do not support embodiment exclusively. ·Femilisk 09:46, 25 May 2014 (UTC)
Go nuts. I freely admit ignorance, being a kind of renaissance polymath with widely scattered interests, not claiming depth of knowledge anywhere in particular. I first got interested in cognitive science when an online acquaintance mentioned Lakoff's WFDT. That, and his "Metaphors We Live By" made so much more sense to me than the classic Greek philosophy that had been previously jammed into my head by academics I viewed as poseurs... I might have become a Lakoff zealot. What with RW's mission being a lot about cutting through bullshit, well, you can probably infer the rest. Sprocket J Cogswell (talk) 14:30, 25 May 2014 (UTC)
No worries, and I don't think embodied stuff is bad just a little on the unfalsifiable side. Think of the whole Lakoff metaphors thing and have a hack at trying to make it testable; derive a prediction from it. ·Femilisk 17:11, 25 May 2014 (UTC)
Thanks for the food for thought... it's brought me back to reading up on the subject. As I understand the Lakoff metaphors thing, "how conceptual metaphors come to be" has to do with the co-activation of neural source and target domains. If I've read it right, it has predicted which topologies of neural modeling more successfully mimic human language learning than others. That might not be the exact plank you're looking to nail down, though.
In faintly related news, of what use is taxonomy? I believe it is coherent, or internally consistent, and taken seriously as a scientific area of study, but what does it predict? Sprocket J Cogswell (talk) 15:51, 27 May 2014 (UTC)
I think the difference is that taxonomy has a clear purpose, is not vague verbal theorising, as far as I know does make predictions (think of all the various missing links, transitional forms, predicted and then found in the fossil record); also no other sensible alternative exists and it accounts for the DNA data well. Not so for embodied theories of cognition! Any way you slice them strong to moderate embodiment does not do the above.
Also embodiment does not predict things over and above other theories. It is unfalsifiable. It is mainly just-so stories. Hacking it up with Occam's razor yields better results (which surprisingly look identical to classical mainstream cogsci). In other words, (strong) embodiment pre-supposes a bunch of things (that the current state of your body is very important). But the brain is doing the exact opposite. It's attempting to abstract over the current state of your body. So unlike a baby you do not forget what you are looking at if you stand up (in the case of a baby the person it is sitting on stands up). Your brain state is not governed by the current modality. If you are running you do not have a different conception of "kettle" vs if you are having a cup of tea. If you think you do, contact me I'd love to test you! :P
I'm not saying all cognition is amodal purely, but saying it's on the spectrum between amodal and modal. And that is what mainstream cogsci claims and not embodiment, which pushes the pendulum all the way towards full embodiment. It's just not possible given the fMRI and behavioural data.
Besides, if you check what I said above it was a two-fold criticism of why this article is problematic. Perhaps embodiment is presented so well, it seems appealing. I suggest you look at some critical works that argue against Lakoff, and against these theories in particular, perhaps? My view is the mainstream view so it won't be hard to find criticism, although I can also find it for you if you don't have access! :) I also really suggest you look at the power and effect size of papers that purport to support embodiment. The stats speak for themselves. This is why stats understanding is so important in science.
I'll leave you with a quote from one of embodiment's main proponents, Barsalou:
Another limitation of current [embodied cognition] work is the relative lack of formal and computational accounts. It is fair to say that current empirical research on grounded cognition heavily reflects demonstration experiments. As philosophers of science note, when a new area emerges, demonstration experiments dominate to justify the area’s importance. Eventually, mechanistic theories develop that stimulate new generations of research, distinguish between mechanistic accounts, and elaborate mechanistic accounts further. Mechanistic accounts of grounded cognition have existed for some time and continue to emerge increasingly (e.g., Cangelosi & Riga, 2006; Farah & McClelland, 1991; Feldman, 2006; Pezzulo & Calvi, in press; Plaut, 2002; Wennekers, Garagnani, & Pulvermuller, 2006). Some preexisting systems have much potential for development as grounded theories (e.g., O'Reilly & Norman, 2002; Ullman, Vidal-Naquet, & Sali, 2002). In addition, various preformal architectures have potential for development as computational systems (e.g., Damasio, 1989; Simmons \& Barsalou, 2003). In general, though, it is clear that much further theoretical development remains, and that such developments will move the area forward significantly.
(Barsalou, 2010, p. 719)
He's trying to sugar-coat it, but it shines though: Mainstream cogsci has mechanistic and functional models, embodiment doesn't. I've tried to create a computational account of embodied theories in my field. It's actually impossible. They end up being classic models, for a number of reasons I won't go into right now as this is getting really long. But that is very damning if your theories are so vague they cannot be made computational. Sorry for the long post! ·Femilisk 16:20, 27 May 2014 (UTC)
If you are running you do not have a different conception of "kettle" vs if you are having a cup of tea. Whait, wat? Who says any different? As I understand it, the neural pathways active when internally visualizing lifting and pouring from a kettle can be identical to the ones used in the factual act, only without the motor activity engaged. Similar things happen in dreams: we walk, talk, and see things, but without the sensory or motor connection to "reality" being complete. Occam's razor would hint that much of the underlying neural mechanism, the identical cells, can be used for both visualizing and doing. There may even be FMRI studies to back that up.
I know for experienced fact that I grok the effect of different chord spellings on something like an Irish bouzouki much better after I've actually grabbed the strings a few different ways and heard the result. After some repetition, I don't have to be holding the instrument for the images to be usable. Similar things have been said about piano keyboards and music theory: by handling and looking at the keys, a player gets a built-in conceptual framework for understanding musical intervals, scales, and chords, and the framework is still available away from the instrument. Same goes for the Japanese abacus (soroban) historically having provided an internal virtual tactile/visual basis for mental calculations.
I suspect we are talking about different notions of embodiment here. If there are those who hold with an extreme form that says immediate bodily state has a strong influence on internal schemas, then I want some of what they're smoking. What I've seen has seemed less like just-so stories and more like the linguists' stereotypical shoe boxes full of note cards, each one a data point, and plenty of them. Sprocket J Cogswell (talk) 17:10, 27 May 2014 (UTC)
OK, you definitely need to examine the spectrum of embodiment. You also need to accept that embodiment is not mainstream cogsci. I've provided you with some references. But seriously, linguistics is dangerously normative. And either way, the linguistics side of cogsci is so a very biased view if not complemented by the four other main branches of the discipline. Is your lab mainly linguistics? ·Femilisk 20:04, 27 May 2014 (UTC) PS: I might be wrong but I'm getting the impression you've not carefully read what I posted. Because one or two of the things I addressed you brought up again. If you have fMRI data that fulfils the criteria I explained above, I'd love to see it. Like I said I work with people who supported embodied cognition until they were tasked to gather unambiguous data. ·Femilisk 20:10, 27 May 2014 (UTC)