Making Algorithmic Systems More Meaningfully Transparent for Users Through Visualisation

Abstract

Transparency over algorithmic systems is increasingly asked for, though it is unclear how it can be realised in practice. Transparency is largely recognised as helpful -eg. to redress information asymmetries -but not as an end in and of itself. Understanding how transparency can be meaningful to users, as well as the challenges and design methods to improve transparency at the interface level is therefore key to make algorithmic systems more usefully transparent for users. As part of the DIS '22 Doctoral Consortium, I am presenting my PhD research on visualising meaningful transparency for users. We are exploring the role of visualisation as a lever to make system transparency more meaningful to users. We consider transparency from a broad socio-technical and systemic perspective, with HCI and Art History research methods. We focus on user interfaces, as they are the key point of interaction between users and systems. By analysing the status quo and various users' transparency needs, we highlight challenges regarding current practices and transparency guidelines, but also ofer practical leads to improve algorithmic system transparency for users. CCS CONCEPTS• Human-centered computing → Human computer interaction (HCI); Visualization design and evaluation methods.

Socially meaningful transparency in data-based systems: reflections and proposals from practice

“…Other research focuses more on the human aspects of meaningful transparency, particularly how appropriate and useful the transparency practice is for achieving stated aims for a target audience in a given context, e.g. Schor et al (2022) and Obar (2022). As Sloane et al (2023) identify, this latter approach defines meaningfulness in the pragmatic tradition as "the property of conveying information that is receivable and useful to a recipient, and that has consequences in that it makes a difference to practice".…”

“…Other research focuses more on the human aspects of meaningful transparency, particularly how appropriate and useful the transparency practice is for achieving stated aims for a target audience in a given context, e.g. Schor et al . (2022) and Obar (2022).…”

“…Our first three and our final two proposals articulate our emphasis on the social in “socially meaningful transparency”, while proposals four and five build on existing individual in context definitions of meaningful transparency in the literature (e.g. Sloane et al ., 2023; Schor et al ., 2022).…”

“…This principle also emerges from some of the literature discussed earlier in the paper (e.g. Pasquale, 2015; Sloane et al ., 2023; Schor et al ., 2022), which argues that what information and how much information to include in transparency accounts are decisions which are as important as deciding which data systems to make transparent.…”

“…In this way, socially meaningful transparency departs from other suggestions for addressing the transparency challenge. It is more socially grounded in nature than, for example, individualised approaches to meaningful transparency (Sloane et al ., 2023; Schor et al ., 2022), or legal approaches such as Obar's (2020) suggestion that individuals could delegate responsibilities to expert practitioners or Pasquale's (2015) idea to make available information that enables individuals to assess the safety of a system.…”