“Will it rain today?”
“How long do I have to wait for my bus?”
“Is the road from the bus stop to my home well lit?”
We are increasingly exposed to sensing and prediction in our daily lives. Uncertainty is both inherent to these systems and usually poorly communicated. To design data presentations that non-experts can understand and take decisions on, we must study how users interpret their data and what goals they have for it. This informs the way that we should communicate results from our models, and visualise qualitative features of the data, which in turn determines what models we must use in the first place.
Visualisation is the actual process of mapping the data to visuals for easy communication. The viewer’s interpretation of a visual is the final stage of visualisation, after which the viewer may decide how to consume the visualisation.
Most viewers consume the visualisation with either of the following two goals in mind:
– gaining new insight into the data represented in the visual, or
– gaining a better understanding of the real phenomena itself.
Often a trial-and-error approach leads to finding the most expressive and effective (graphically articulate) visualisation. Yes, the trial-and-error design process involves developing the visualisation in accordance with the already established theory and principles and user study with an iterative design process where the actual user is kept in the loop.
However, the value of a visual for the purpose of a particular interpretation is not obvious to the viewer before its use for interpretation. The same visual might bring about new insights to one user, but not to another; the same visual might be effective for one problem, but not for another; the same animation might be adequate to understand a problem on one type of hardware, but not on another.
In order to generate the most meaningful visualisation for a specific instance, a careful mapping process from “data to visuals” is necessary. And it will vary a great deal depending upon the preconceived knowledge of the users, their mental models, and the design of the visualisation, among other factors.
The “user model” describes the collective information the system has of a particular user.
A visualisation is subjectively interpreted by the viewer in dependency of past experiences, education, gender, culture, situation, and individual limitations, abilities, and requirements. For instance, colour-deficient viewers are limited in interpreting colour pictures; a person with deficient fine motor skills will have problems accurately pointing at small objects on the screen. In order to create a user model, the system needs to learn facts about the user. Most of these facts can be extracted from observing the user perform special tasks.
A complete user model evolves in several stages, whereby each style of user modelling is being used. Typically, the extraction of information starts with explicit modelling to inquire about gender, age, or education. Subsequently, the user has to complete special tasks that reveal the limitations of his/her vision and/or preferences. By continuously observing the user in his/her use of the visualisation system, the user model can be improved over time. Significant information of the user model is expected from the completion of special tasks.
I work on a research project titled, ‘Deep User Models for Visual Analytics’. With an aim to understand how to communicate the uncertainty to non-experts who have no technical background and also at the same time maintaining the relevance of the project for domain experts we built our first study around the public transport in Melbourne, Australia. Through this project, we are trying to understand the Perception of Visual Uncertainty Representation by Non-Experts. The motivation is that understanding and communicating uncertainty and sensitivity information is difficult; uncertainty is part of everyday life for any type of decision-making process, and some of the previous studies done are unclear and could be improved.
The question we tried to answer is: Can we build visualisations of uncertainty distributions (specifically, public transport arrival times) that people understand? More specifically, our study investigated whether a particular visualisation of uncertainty information in predicting the arrival time of one bus and the departure of another could be used to help people make a transfer, but will involve a more complex visualisation.
We are looking at how to tune models to people’s error preferences in a simple, lightweight way. It is not enough to add an effective visualisation on the existing models. Even an effective representation of uncertainty, in this case, might not be optimal if the model is not tuned to reflect people’s error preferences. Given known costs for each type of error, cost-sensitive classification can be employed to fit a model that makes predictions that reflect error preferences.
Our first study related to designing a transit mobile application for public transport in Melbourne, tries to help commuters make transfers among various modes (bus, train, and tram) by making use of visualisation to communicate the associated uncertainty in arrival and departure times. The findings from this study will help design user facing applications can leverage the power of visualisation to communicate uncertainty information to non-experts.
Our planned second study, will try to build user models in an attempt to understand how non-experts and experts perceive visualisations in their daily life. The findings from this study will help us come up with guidelines for designing visualisations that people can understand and then take effective decisions.
We, graduate research scholars of the IITB-Monash Research Academy, study for a dually-badged PhD from IIT Bombay and Monash University, spending time at both institutions to enrich our research experience. The Academy is a collaboration between India and Australia that endeavours to strengthen relationships between the two countries. According to its CEO, Prof Murali Sastry, “The IITB-Monash Research Academy was conceived as a unique model for how two leading, globally focused academic organisations can come together in the spirit of collaboration to deliver solutions and outcomes to grand challenge research questions facing industry and society.”
He is right! Visualisations are often targeted for experts in a domain. I have always been fascinated by how a good visualisation design can help us understand the underlined information, trigger an emotion, and guide us in taking an informed decision. This project offered me a chance to develop a deep understanding of how visualisations are perceived by the people. This will guide the designer leverage the power of visualisations to communicate complex phenomena to people.
Research scholar: Amit Jena, IITB-Monash Research Academy
Project title: Deep User Models for Visual Analytics
Supervisors: Prof. Venkatesh Rajamanickam, Prof. Tim Dwyer, Dr. Ulrich Engelke, Dr. Cecile Paris
Industry Supervisors: Dr. Ulrich Engelke and Dr. Cecile Paris, Data61 CSIRO
Contact details: firstname.lastname@example.org
The above story was written by Amit Jena. Copyright IITB-Monash Research Academy.