Action without Interaction

Guideline: Use interaction in visualization sparsely and cautiously.

I accidentally found this guideline online. I guess that the author (Gröller) may not wish to call it a guideline, but I find this intriguing as many in visualization seem to argue for more interaction. I also find this thinking may be useful in many situations. I am wondering how much interaction is too much and how much is too little. I also wonder if the amount of interaction is a function of data, users, and tasks. Can Der Meister or his former and current students offer more insights on this guideline?

Interaction typically requires considerable effort and cognitive load. Static reproductions have to do without. So there must be very good reasons to resort to interactive investigations. If interaction is necessary, the user should be supported by, e.g., guided suggestions.

1 Like

I think that it is useful to distinguish between visualisations that require user interaction in order to be useful at all, and those that merely allow the user to interact.

It is possible to have a visualisation that works as a static visualisation but also allows the user to further explore using interactivity, such as by mousing-over a data point to see the precise value corresponding to that point or additional information about it in a tooltip.
I think that there is generally little cost (to the user) associated with adding such interactivity to a visualisation, as long as it is done in a way that does not create clutter or confusion.

One of the issues mentioned in Groller’s slides is that “interaction hampens reporting”. I think that the transfer of viewer state for interactive visualisations in order to facilitate collaboration and handoff between users requires further work.

1 Like

To follow up on this interesting discussion, we conducted a user study with a hypothesis – “the amount of interaction may influence a person’s opinion as how easy it is to use a visualization tool”. If the same task is performed using two different tools, the tool that requires fewer interactions or less interactive time corresponds to “using interaction” more “sparsely and cautiously”.

The study consists of four trials. For each trial, participants are required to create two bar charts to visualize a given data set. We use MS-Excel and Tableau as the two visualization tools. At the end of the second and fourth trials, participants were given s subjective questionnaire to rate the easiness of using the two visualization tools and for the overall impression. We collected – the number of interactions (NI) and the task completion time (CT).

We observed that that participants’ opinions on the relative easiness depend mainly on the task completion time, and to some extent also on the number of interactions. In some cases, the familiarity of the software may modify the direct correlations among NI, CT, and subjective judgement.

The data for the experiment can be found at GitHub.