What does it mean to trust a visualization?

Exploring the dimensions of trust perception in InfoVis

Few concepts are as ubiquitous in computational fields as trust. Whether it is a robot’s behavior, the software solution’s outputs, or the machine learning model’s reliability, trust is a hot topic in any situation where humans interact with machines. This is no less the case in the field of Information Visualization. However, in the case of InfoVis, there are several unique and complex challenges, chief among them: defining and measuring trust. Therefore, we posit that “trust” has the most utility when considered a catch-all term for other factors.

Clarity

Borgo \& Edwards (2020) claim that there are three “bases” of trust: ability, benevolence, and integrity. Applying these terms to visualizations is not immediately apparent, so the core metrics behind them must be fleshed out. First is ability. When one claims that a trustee is “able,” it is assumed that the trustee can do the task entrusted to them. We can translate this for visualization as the readability or clarity offered by the visualization, as communicating information is arguably the most crucial role of data visualization.

Credibility

The second is “benevolence.” When one claims that a trustee is “benevolent,” it is assumed that the trustee will act in the best interest of the trustor. In the visualization field, this can be understood as the credibility of the visualization. In other words, the expectation that the visualization is not maliciously or incompetently created in such a way as to lead the audience to a false conclusion.

Reliabilty

Lastly, there is integrity. When one claims that a trustee has “integrity,” it is assumed that the trustee will adhere to some behavior or principles. In other words, the trustee is reliable and will perform the tasks expected. Visualizations can be seen as the perceived ability of the audience to rely on or use the information given, especially in decision-making.

Summary

In summary, these three aspects of trust laid out in prior work can be repurposed for visualization as clarity, credibility, and reliability.