Knots are fascinating: they tie together topology, embodied experience, and material culture. As Thai textile artist and designer Nithikul Nimkulrath (2024) has pointed out, knots are the kind of thing we come to know “through and in making”. One of her artworks (see photo) explorers the materiality of knots by translating a physical, hand-knotted container into a 3D render.
As feminist approaches have long realized, knotting, like knitting, is also one of those fiber crafts that have always featured a pleasing and compelling entanglement of craft and computation. Knotting is an “embodied approach to engaging with the world”, as Vasiliki Tsaknaki and colleagues describe in a recent paper. One of my favourite freewheeling authors in cognitive science has described knots as “a wonderful case study for embodied and object-dependent cognition, as their realization depends on continuous object and sensorimotor feedback” (Casati 2013).
So I found myself equal parts intrigued and irked by this fun Open Mind article on knots, by Sholei Croom and Chaz Firestone. Intrigued, because knots are often counterintuitive and reasoning about them is hard (try explaining how to tie a shoelace). Mildly irked, because the paper somehow manages to say nothing about knotting as an embodied process — instead defining knots as “naturalistic stimuli that appear across cultures”.
How do they “appear”? When are they ever just “stimuli”? Why should we expect people to reason about them by just looking?
To be fair, those are not the kind of questions the paper sets out to address. Instead, it positions itself as a contribution to debates on intuitive physics, where researchers have hypothesised a mental “intuitive physics engine” that can be studied by asking people to reason about abstract representations of physical situations like falling blocks or moving balls. So this paper uses pictorial representations of knots and asks people —in a total of four experiments— to reason about them.
Poor performance
Experiment 1 uses static pictures of handknotted specimen of the kind shown in Figure 1A,B. Discussing the results —in brief, people are unable to reliably predict knot strength merely by looking at pictures, no surprise here— the authors note: “it is possible that poor performance was driven by auxiliary factors that prevented subjects from accessing or demonstrating subtler and more accurate physical knowledge”.
Here I perked up. How are the researchers going to deal with those pesky “auxiliary factors” that somehow prevented people from accessing “subtler knowledge”? Maybe the next set of experiments will let folks interact with actual physical knots, the way they do everyday?
Nope, that wouldn’t be intuitive physics (also, it would be hard to administer online). Experiments 2-4 are basically the same, but with a larger variety of 2D images, now derived from 3D renders. Why renders? The paper muses that perhaps there were “inadvertent consistencies” in the hand-knotted samples; or perhaps it was because the pics in study 1 featured “only one perspective” and made it hard to “extract topological organization”.
(I really felt that at this point, the authors were… stringing us along, pardon the pun. Did they seriously think it was anything like those factors and not quite simply the fact that folks were unable to experience and explore the knots in the ways they normally do, by feeling, tugging, picking and pulling at threads?)
Intuitive physics sounds like a cool field for thought experiments and mental gymnastics, but using knots to study it, particularly pictures of knots, seems far from intuitive to me.
Why expose people to visual representations that don’t capture anything of our mundane embodied experience with knots? Which climber would ever judge a knot by mere visual inspection? Who even looks at their shoelaces?
Across all four experiments, the paper finds that people are basically unable to reliably judge the strength of a knot by visual inspection alone. They are fairly consistent, but end up being consistently wrong and consistently right in equal measure. None of this will be surprising to anyone who has ever entrusted their life, their climbing partner, or their gear to a knot.
A Trojan horse
The paper’s main conclusion —”knots strain our reasoning”— seems to assume a strictly skull-bound view of reasoning that treats knots as sui generis objects and people as disembodied mentalizers. When it comes to intuitive physics, apparently we might as well be a brain in a vat.
And yet, and yet. One way to read the results is as suggesting that mere visual exposure is not enough to ground at least some kinds of reasoning. This could point to a key limitation for so-called “general purpose” models of scene understanding — most prominently in times of AI hype, multimodal machine learning models that subsist on nothing but pictures and words. That rescues at least some of the relevance of this paper for me.
We might even take this a couple steps further. It seems quite remarkable for a paper “revealing a blindspot in theories of physical reasoning” (final paragraph) to fail to provide any directions for resolving this gaping hole. In fact, if experimental evidence is your thing, this set of studies could be a clinching argument, in photo negative as it were, for embodied, enactive, embedded, and extended accounts of human cognition — amazingly, without mentioning any of those notions.
You wouldn’t think so by merely looking at it, but after pulling at some of its threads, turns out this paper is a Trojan horse for 4E cognition.
References cited
- Casati, Roberto. 2013. ‘Knowledge of Knots: Shapes in Action’. In Shapes 2.0: The Shape of Things, edited by Oliver Kutz, Michel Bhatt, Stefano Borgo, and Paulo Santos.
- Croom, Sholei, and Chaz Firestone. 2024. ‘Tangled Physics: Knots Strain Intuitive Physical Reasoning’. Open Mind 8 (September): 1170–90. doi:10.1162/opmi_a_00159.
- Nimkulrat, Nithikul. 2024. ‘Traces of Craft Experience in Artefacts’. In Craft and Design Practice from an Embodied Perspective, edited by Nithikul Nimkulrat and Camilla Groth, 141–54. New York, NY: Routledge. doi:10.4324/9781003328018.
- Tsaknaki, Vasiliki, Lara Reime, Marisa Cohn, and Tania Pérez-Bustos. 2024. ‘Knotting Data as a Feminist Approach to Data Materialization’. Edited by C. Gray, E. Ciliotta Chehade, P. Hekkert, L. Forlano, P. Ciucarelli, and P. Lloyd. DRS2024: Boston (DRS Biennial Conference Series), June. https://dl.designresearchsociety.org/drs-conference-papers/drs2024/researchpapers/270.