Hi Eiko! Will you be attending AIMOS? If so, I'd love to catch you during a break and perhaps chat all things measurement schmeasurement / jingle-jangle and the role of formal models in building theories in psychology. I have a few favourite cafes in Sydney too! 🙏

For the Feyerabend anarchists:

me: finding a recent and relevant Chaz phil vis paper that I had missed through zohran-posting

I'll be presenting this work at the Australasian Society of Philosophy and Psychology conference (and alluding to it at the Australasian Cognitive Neuroscience Society conference) #ACNS. Would love to chat to anyone interested about this work! Please feel free to reach out! /fin

So, what is a working memory representation? More from me to come on that, including further development of these models, a couple of #cogneuro projects in prep, and a follow up on my "theory map" piece. But for now, please check out our preprint: osf.io/fm9vz /10

We think we have provided a principled modeling approach that can clarify the interactions of stimulus-specific effects and more general mechanisms in VWM. Our approach can help #cogsci better define cognitive models of representations and mechanisms as explanations of task performance. /9

We also find that the recovered representation is improved by refining the cognitive mechanisms (here, swap errors in memory reproduction). This emphasizes a key point – that the representation should be incorporated in cognitive models, not simply offloaded to similarity judgments. /8

As per Schurgin et al. (2020), we find that the cognitive representation underlying all tasks do not match the physical stimulus space. But we also find that the representation is different for similarity comparison and both reproduction tasks; similarity is not the basis for working memory. /7

Our models fit for the cognitive representation jointly with the theorised mechanisms using the task data itself, rather than treating the representation as independent to the model (or assuming that the psychological representation is similarity-based). The model code can be found here: /6

We fit models for three separate tasks: a similarity comparison task, a perceptual reproduction task, and a working memory reproduction task. The dataset comes from Tomic and Bays (2024): doi.org/10.1037/xlm0.... Thanks to @ivntmc.bsky.social and @bayslab.org for providing the open dataset. /5

Recent work has found adjustments for psychological similarity using an approximation or empirical measures does not lead to better model fits of VWM performance. Oberauer (2023): psycnet.apa.org/doi/10.1037/...

But the question remains, what is the cognitive representation underlying VWM? /4

The TCC model is like many other formal models of cognition that collect similarity judgments to then use multidimensional scaling methods and derive the psychological representation. The cognitive mechanisms are built upon that MDS representation to predict the cognitive task performance. /3

The issue of defining the appropriate representation is core to the field of visual #workingmemory. Schurgin et al. (2020) introduced the TCC model and influentially argued that models of VWM should be in terms of psychological similarity, rather than in terms of the physical stimulus space. /2

A bit of a stretch but how about SoundCheck? Mainly for the pun, could imply research that has open materials is more sound...