
Dr Yi-Shin Lin
- Position: Research Fellow
- Areas of expertise: Cognitive Psychology; Mathematical modelling; Behavioural research methods; Bayesian inference; Static and dynamic stochastic differential equations. Software engineering. C++, CUDA, R and Python.
- Email: Y.S.Lin@leeds.ac.uk
- Location: Room 8.238 Driving Simulator, Physics Research Deck
- Website: Software | Notes on Cognitive modelling | Twitter | Googlescholar | ORCID
Profile
I am a cognitive psychologist specialising in behavioural decision-making in applied contexts bearing significance for society. My recent interests focus on using decision-making in neuro-behavioural economics in the choice behaviours in traffic and decarbonisation, such as choosing electric versus other vehicles by bringing in the knowledge of high-performance quantitative modelling.
Responsibilities
- Research fellow in Human and AI cognition in COMMOTIONS project
Research interests
My research with the Human Factor and Safety (HFS) Group focuses on road safety. Together with colleagues in HFS, we study road user behaviours, mainly how an interaction might evolve into traffic collisions. In one study, we investigated the lane-merging behaviour in drivers by applying a machine-learning model dubbed the Long Short-Term Memory (LSTM) recurrent neural network (Srinivasan et al., 2021; 2023). This finding goes hand-in-hand with our new cognitive model that applies the basic principle of utility maximisation, discrete motor primitives, and sensory evidence accumulation to explain pedestrians' and drivers' behaviours in traffic interaction (Lin et al., 2022).
My research in the last few years spans human vision, decision-making, Bayesian computation, and computational modelling. I began my PhD project with interest in examining the visual search process and designed a computational model to analyse the distribution data in the two-choice visual search task. I then studied how the Bayesian modelling might help to explain the distributional data. This work resulted in a hierarchical Weibull distribution model to fit the joint response time and response choice data in a series of visual search experiments. This work contributes to the field by showing that one can infer the processing of visual search by examining different bands (e.g., fast vs slow RTs) in the RT distributions. I published this software on GitHub and documented it in the paper "Modeling visual search using three-parameter probability functions in a hierarchical Bayesian framework" (2015) when the open science movement was still in its early time.
While with the Tasmanian Cognition Laboratory, I led two projects developing new statistical methods using Bayesian computation. In the first project, I designed a new version of the complex decision-making model, the Piecewise Linear Ballistic Accumulator model. This version harnesses parallel heterogeneous computation using GPU-based CUDA and the C++ language. I also created a tool to fit complex decision-making models which were previously unfeasible. I have also shared the result of this work on my GitHub and documented its details in the paper "Parallel probability density approximation" (2018). My second project was a generic solution for conducting hierarchical Bayesian modelling. I implemented one Bayesian sampling method, population-based Markov Chain Monte Carlo (MCMC), which is particularly effective in cognitive models featuring correlated parameters. The source code is also available on my GitHub. I summarised this work in a paper titled "Evidence accumulation models with R: A practical guide to Hierarchical Bayesian Methods".
I have contributed to several projects with the data and code archived in the OSF. You may find them helpful.
https://osf.io/av3cm/
https://osf.io/p4pdh/
https://osf.io/pbwx8/