iiinova — where your digital self keeps evolving

Jake Quilty-Dunn @quiltydunn.bsky.social · 23/09/2026

mcdowell has my favorite instance of this

"I do not mean to be objecting to anything in cognitive science."

Jake Quilty-Dunn @quiltydunn.bsky.social · 13/09/2026

our main claim is that there is no coherent level-independent N vs E debate, as we tried to say in the final paragraph here. i would argue that your points here actually support this idea - the tractable debates are localized to particular levels, that marcusian 'nativism' =/= fodorian 'nativism'

110 0

Jake Quilty-Dunn @quiltydunn.bsky.social · 11/09/2026

There is a lot more in the paper, including how domain constraints shape attractors that brains and artificial models settle into + the role of process-based vs stable-"thing"-based orientations in science. We also sketch ways forward for developmental science once the N vs E debate is dissolved.

140 0

Jake Quilty-Dunn @quiltydunn.bsky.social · 11/09/2026

We propose a multi-level framework: 1. Subsymbolic 'learning' operates below the level of rational, psychological operations, tuning parameters to minimize loss. 2. The brain thereby enters into attractor states in this neural state space. 3. Some attractors realize symbolic representations.

image of empiricist and nativist saying an early behavior by chick is learned and innate, respectively, with diagram of multilevel framework for learning

140 0

Jake Quilty-Dunn @quiltydunn.bsky.social · 11/09/2026

We review a recent literature, including work done by Justin's lab, that tightly controls the inputs available to developing animals and to computational models or "digital twins". I think this is some of the most fascinating work being done in cognitive science right now.

Diagram of digital twin studies

1. Raise newborn animal in controlled environment (training phase)
2.Test what animal learned in the environment (test phase)
3.Build populations of artificial agents (brain & body models)
4. Raise agents in animal-matched environment (training phase)
5.Test agents on animal-matched tasks (test phase
6.Compare development of animals and agents (evaluation)

120 0

Jake Quilty-Dunn @quiltydunn.bsky.social · 30/08/2026

idk if we have absorbed it fully though. i always thought lovecraft nailed this:

The most merciful thing in the world, I think, is the inability of the human mind to correlate all its contents. We live on a placid island of ignorance in the midst of black seas of infinity, and it was not meant that we should voyage far. The sciences, each straining in its own direction, have hitherto harmed us little; but some day the piecing together of dissociated knowledge will open up such terrifying vistas of reality, and of our frightful position therein, that we shall either go mad from the revelation or flee from the deadly light into the peace and safety of a new dark age.

031 0

Jake Quilty-Dunn @quiltydunn.bsky.social · 21/08/2026

the APA is organizing these too far in advance imho

An email about the APA central meeting in the year 2207

0102 0

Jake Quilty-Dunn @quiltydunn.bsky.social · 14/08/2026

strange that it's listed as solo authored, by him, on his website and CV in that case

150 0

Jake Quilty-Dunn @quiltydunn.bsky.social · 17/06/2026

I also argue that some of these constraints can be loosened, allowing for a hierarchy of more or less narrowly iconic representations. Representations that are "higher" on the hierarchy might be widespread in the mind/brain, perhaps underlying higher forms of cognition.

sketch of hierarchy of icons, from individual analog dimensions to coordination into a multidimensional space (e.g., color), to coordination of dependent with independent dimensions (e.g., some graphs, diagrams), finally to coordination of dependent dimensions with monogamous, spatiotemporal independent dimensions (e.g., photographs, visual icons).

030 0

Jake Quilty-Dunn @quiltydunn.bsky.social · 17/06/2026

For Haugeland, independent dimensions are special bc they are (what I call) "mandatory": every value along the dimension must be tokened in the icon. Dependent dimensions are optional. Every spatial location must be represented, but not every altitude or color value must be.

120 0

Jake Quilty-Dunn @quiltydunn.bsky.social · 17/06/2026

Borrowing some concepts from John Haugeland, I argue that icons coordinate values along "dependent" dimensions (like color, or altitude in an altitude map) with "independent" dimensions (like spatial dimensions in both photographs and maps).

130 0

Jake Quilty-Dunn @quiltydunn.bsky.social · 17/06/2026

Scene-based early vision contrasts with object-based vision, which segments objects, tracks them over time and space, stores their features, and makes each object available to visual working memory. Intuition pump: scene-based vision thinks there are curved lines in this image

illusion where you can see curved lines but only when you're not looking directly at them

120 0

Jake Quilty-Dunn @quiltydunn.bsky.social · 17/06/2026

The candidate class of visual processes are what I call "Scene-based Early Vision": a suite of processes that integrate the visual scene, detect the presence of features, briefly store the distribution of features, summarize them, and make those summaries available to visual working memory.

110 0

Jake Quilty-Dunn @quiltydunn.bsky.social · 11/06/2026

me watching the fourth quarter of that game last night

091 0

Jake Quilty-Dunn @quiltydunn.bsky.social · 20/03/2026

Our commentary cites evidence from the past 15 years of research on object continuity in vision science demonstrating that the famous Spelke criteria for objecthood - including spatial continuity and cohesion - are neither necessary nor sufficient for object continuity in visual perception.

figure illustrating experimental paradigms used to show that cohesion and spatial continuity are neither necessary nor sufficient for perceived object continuity

110 0

Jake Quilty-Dunn @quiltydunn.bsky.social · 15/03/2026

scenes from the MIT philosophy department

3265 0

Jake Quilty-Dunn @quiltydunn.bsky.social · 08/11/2025

3) how should we think about old psychology today? one lesson from the replication crisis of 10 years ago is that effects are probably much smaller than the older work made it seem. aside from in-your-face phenomena like visual illusions, the effects are probably not huge (e.g., semantic priming)

151 0

Jake Quilty-Dunn @quiltydunn.bsky.social · 08/11/2025

the authors concede in a reply to pauer et al. that "skepticism about our [null] results is warranted given the findings related to perceived choice." so: this was a positive replication of 2 dissonance effects, with some hard-to-interpret data regarding choice. far from the end of dissonance

120 0

Jake Quilty-Dunn @quiltydunn.bsky.social · 08/11/2025

so perhaps this was not really a low-choice condition, but "medium-choice". then, lishner and pauer et al. analyzed the data and found that, despite the lack of difference between conditions, increases in perceived choice across and within conditions significantly correlate with attitude change!

analysis from lishner 2024's commentary: "Also notable is an internal analysis based on the data provided by Vaidis et al. (https://osf.io/9xsmj), which revealed a significant correlation between ratings of perceived freedom and attitude change among participants within and across both counterattitudinal conditions (across both conditions: r = .12, p < .001, N = 1,997; low-choice condition: r = .17, p < .001, N = 937; highchoice condition: r = .09, p < .003, N = 1,060)

120 0

Jake Quilty-Dunn @quiltydunn.bsky.social · 08/11/2025

but the "low-choice" participants did not indicate that they felt unfree: instead, they indicated a mean of 4.44 on a scale of 1-9 (9 being the most free), and "high-choice" had a mean of 6.50. as the authors note in a follow-up paper, this is higher than "low-choice" in *over 90%* of studies (!!)

vertical black line showing the level of choice in the "low-choice" condition, which was around 50%, and higher than over 90% of all other studies sampled

120 0

Jake Quilty-Dunn @quiltydunn.bsky.social · 08/11/2025

croyle & cooper originally found that choice did matter. vaidis et al. did not replicate this: high-choice and low-choice were the same. however, this was not because there was no evidence for dissonance effects, it was rather because dissonance effects were observed for *both* conditions

120 0

Jake Quilty-Dunn @quiltydunn.bsky.social · 08/11/2025

so doubts about the scientific standards of "When Prophecy Fails" are not new, or especially relevant to dissonance theory. it's not very different from concerns re: the story about ben franklin flying a kite in a lightning storm; relevant for history, not so much for the study of electromagnetism

19th century painting of ben franklin discovering electricity

181 0

Jake Quilty-Dunn @quiltydunn.bsky.social · 06/11/2025

the lack of effect of choice is completely consistent with festinger's original theory and some more contemporary versions of dissonance theory, as the authors explicitly say. it is an interesting null result but it is not a troubling one for the core claims of dissonance theory.

1101 0

Jake Quilty-Dunn @quiltydunn.bsky.social · 06/11/2025

josh, this is also misleading. they successfully replicated the two central claims of dissonance theory: attitude change, and increased negative affect. the latter is inconsistent with virtually every extant alternative to dissonance theory.

3142 0

Jake Quilty-Dunn @quiltydunn.bsky.social · 25/09/2025

it's beautiful work, and it's exactly the sort of thing we had in mind as an alternative to model-based planning. rewarding randomly generated activity to build a route in a spacetime attractor network is v different from using an internal model to simulate counterfactuals and select among options

141 0

Jake Quilty-Dunn @quiltydunn.bsky.social · 24/09/2025

nyc baby, greatest city in the world

0110 0

Jake Quilty-Dunn @quiltydunn.bsky.social · 22/09/2025

We argue that the data do not support this (surprisingly common) interpretation. Hippocampal replay is often spatially fractured and physically unrealizable, and lacks the small-scale dynamics that would be necessary to support model-based planning. Other functions are more plausible. (3/7)

Figure 2 from McNamee, D. C. (2024). The generative neural microdynamics of cognitive processing. Current Opinion in Neurobiology, 85, 102855. The figure describes multiple functions of hippocampal replay without positing model-based simulation.

161 0

Jake Quilty-Dunn @quiltydunn.bsky.social · 22/09/2025

Many researchers think that cognitive maps in the hippocampus are used for implicit simulations in humans and non-human animals, both during waking states and sleep. This view is based on activation patterns that correspond to spatial information and aid future navigation. (2/7)

Figure 2 from Carr, M. F., Jadhav, S. P., & Frank, L. M. (2011). Hippocampal replay in the awake state: a potential substrate for memory consolidation and retrieval. Nature neuroscience, 14(2), 147-153. The figure shows rodent hippocampal sharp wave ripples during navigation in awake replay, both local and remote.

141 0

Jake Quilty-Dunn @quiltydunn.bsky.social · 19/08/2025

this is from the paper. it's also important to think about what the actual plausible rival explanations of the effects are; it's hard to see how an orientation-based story, e.g., can capture the real-world size effects

Of course, even containing the xaet same pixels does not (and eould not hold every mid-and
low-level feature constant For example, rotating the anagrans changes the orientation of edges
in the stimulus, and can also change an mage's aspect-ratio. It would, of course, be impossible to
hold all mid- and low-level features constant without using the exact same image (without
transformation); but in that case, there would be no effects to discover ( barring contextual or
attentional effects). In other words, while this approach does not allow complete control of mid-
and low-level features, it allows for more control than has previously been possible while
systematically altering the high-level property of interest.

230 0

Jake Quilty-Dunn @quiltydunn.bsky.social · 09/08/2025

more random exclamation points in working memory papers please

Although this approach helps to control for uncontrolled influences of past knowledge and skill with the memory materials, it fails to acknowledge the transformational importance of past knowledge and skill with the memory materials!

050 0

Jake Quilty-Dunn @quiltydunn.bsky.social · 02/08/2025

i think core cognitive representations are supposed to be concepts in one sense (they have a certain kind of inferential role and abstract content), but not fully concepts in another sense (they have a different format from, and are more closely tied to perception than, lexical and logical concepts)

from 2011 BBS precis of The Origin of Concepts:
"Core cognition resembles perception in many respects that distinguish it from other types of conceptual representations. These include the existence of innate perceptual input analyzers that identify the entities in core domains, a long evolutionary history, continuity
throughout development, and iconic (or analog) format. The representations in core cognition differ from perceptual
ones, however, in having conceptual content."

130 0