Perceptual biases and visualisations of data

 

Lab rotation project description

This project will introduce you to working with multivariate data (concepts, ideas for organising data). You will also be working on methods for efficient statistical plotting, with a view to best visual encoding of information in the different aspects of plots and data visualisations.

You will get the opportunity to get hands-on experience with R and Julia (an open-source, scientific computing language). You will then work on creating high-quality data visualisations using two different libraries (R/ggplot2 for static plots and vegalite.js for interactive visualisations).

For neuroscience / psychophysics background, we will also introduce you to measurements of thresholds (perceptual limits) and biases (perceptual distortions) using standard methods used in our field. Stimulus presentation software is already setup, but you will have the opportunity to learn about model-fitting (using non-linear least squares) to estimate key summaries of behavioural performance.

Fact file

Research theme

MCO

Location

Add location here...

Rotation

Add rotation details here...

Contact

2nd supervisor


BBSRC Doctoral Training Partnerships
 

Linked PhD Project Outline

This project will establish how and when human perception of data plots is biased. In collaboration with data journalists at the Financial Times we will design individualised data visualisations that overcome these perceptual distortions and minimise confusion and inconsistency across different observers.

Understanding how to best present complex information in plots and data visualisations is important across a wide range of settings: from business to media, from scientific research to public policy making. But creating plots that convey the correct meaning of the underlying data is not trivial. An observer's understanding of what plots and graphs represent relies on visual perception which can be markedly biased by simple changes like adding lines or changing the colour scheme. Many other factors can also influence what people perceive: what the person looked at just before, what they expect to see, and what they are asked to do with the information that is shown to them.

Here, we will rigorously test the hypothesis that perceptual biases distort the interpretation of data visualisations. Lab-based, psychophysical methods provide a powerful and quick way to bridge this gap and build a detailed understanding of the rules that govern perceptual decisions for a wide range of graph types and encodings. To understand how these principles apply to larger populations, we will use online studies and interactive experiments co-organised with our collaborators at the Financial Times, including the popular chart doctor series. 

 

Biotechnology and Biological Sciences Doctoral Training Programme

The University of Nottingham
University Park
Nottingham, NG7 2RD

Tel: +44 (0) 115 8466946
Email: bbdtp@nottingham.ac.uk