Ethics of Dynamic Adaptation to Domain Situations

#1

Guideline: Know Your Audience

The approach, to design data visualization specifically targeted to a group of users, is clearly powerful.
And the manual application of this guideline not only allows for a careful consideration of all implications but also clearly show limiting gaps of our understanding of the end-user.
As a developer of big data tools I find these gaps to close rapidly. With the continued research into data visualization and with the increasing accuracy and power of big data tools, it seems inevitable that in the future we will have dynamically adapting visualizations that allow to target a multitude of users simultaneously.

As a sceptic I must also concede that our tract record of getting things right the first time around (or how ever many iterations the models need) is quite poor. And in contrast to the optimization of ad-buys, the implications of a systematic bias to the access to data is ethically very concerning. Yet it has been quite difficult to find fleshed out guidelines in neither the big data nor the data visualization fields.

Question:
Is it ethical to make use of big data tools that enables us to dynamically adapt visualization designs in order for us to"better" target the person accessing our service?

#2

There are a few research papers on “dynamically adapt visualization designs”, e.g.,
Wongsuphasawat et al. Voyager: Exploratory Analysis via Faceted Browsing of Visualization Recommendations, TVCG, 2016 (https://ieeexplore.ieee.org/document/7192728/)

Healey et al. Visual Perception and Mixed-Initiative Interaction for Assisted Visualization Design, TVCG, 2008 (https://ieeexplore.ieee.org/document/4359504/)

Cook et al. Mixed-initiative visual analytics using task-driven recommendations, IEEE VAST 2015 (https://ieeexplore.ieee.org/document/7347625/)

Most of these papers are about visualization systems designed for observational and analytical visualization tasks rather than disseminative visualization tasks. The goal is to reduce the effort and cognitive load of the users. The users’ knowledge about the tasks and the background of data alleviate the potential biases that may be caused by a visual representation recommended by an algorithm.

You are right. In terms of disseminative visualization, there is an ethical question for safe guarding against any potential negative effects of dynamically adapt visualization designs. Further research on this topic will be necessary for this important topic.

In general, any blackbox algorithms for making decisions for the users in the name of “this is best for you” has potentially an ethical question. In most of such cases, interactive visualization can enable users to explore different options and alleviate the biases of the algorithms.