Session 5: Accessible data visualizations + data journalism

- | #dataviz | Accessibility Club Meetup #8

Session notes

  • Author: Sebastian Greger

    Starting point: a Berlin meetup presenting recent interactive data journalism projects triggered a discussion about the (limited) accessibility of data journalism that is based on visual concepts Example for discussion: map of germany with datapoints - how do we make it accessible? https://interactive.zeit.de/strassennamen/

    • concept is based on a geo-referenced database of all german street names; in addition to a journalistic piece highlighting some of the found insights, this tool allows every reader to enter their street name and see its geographical prevalence all over germany

    this could even be of legal relevance, e.g. for public broadcasters?

    relation to alt text issue: what is the right thing to write into the alt text (good canonical resource: https://axesslab.com/alt-texts/)

    with interactive data visualisation pieces, we are talking about "largely insemantable a11y issues"

    • have a piece of code that analyses the data into text
    • are we misleading when we are translating the data for the user instead of letting them frame them by themselves?
    • ...

    for simple visualisations, e.g. bar charts, svg makes alt texts obsolete

    for the particular street name map example:

    • amount of data points is overwhelming
    • interactivity makes it impossible to have a textual description of every possible view

    one of the inclusive design principles (ref. https://inclusivedesignprinciples.org/): "offer choice"

    • BBC: design principle for every complex vis to provide two panels: one = summary, second = data view
    • hidden table for screen readers: it might also be relevant to others users (not using assistive tech), at the same time it might be too big to comprehend for screen reader users
    • idea from discussion: download button as part of the a11y strategy

    idea: give people possibility to skip over the data (e.g. make it available in a hidden panel)

    ask yourself: what is the story you want to tell?

    • don't kid yourself that it is entirely neutral
    • even journalism itself is never neutral, not to speak of visualisation algorithms or ai algorithms

    Example of an accessible (yet much less complex) visualisation: https://worldwideweb.cern.ch/timeline/ (multiple parallel timelines whose relation to each other is represented differently in various screen sizes) - "this was also designed to work on the original NeXT browser"

    the question of interactivity

    • give people means to expose the connections themselves
    • in the CERN example: ripple lines could be shown when hovering to make paralellity of timelines more obvious; these would however be impossible to translate into a screenreader-friendly format
    • aria-labeled-by can be used for certain interactions
    • "live regions" are also a tool for the purpose, but come with usability issues as screenreader users are not used to them; they believe the focus has jumped somewhere else instead
    • events on focus are an issue

    ai: can i hand it over to a service?

    • "this could be the only context to imagine where there is an actual application for ai"
    • have ai do the interpretation on the user's behalf
    • raises at the same time the usual issue of algorithm transparency/biases

    map implementation on https://thesession.org/members/1 - designing the correct alternative representation of the map was not easy

    • just the geo coordinates are not a useful output for humans
    • problem of duplicating the information
    • uses a geo api to output an approximate location (e.g. town or neighbourhood)
    • "location is political": even just naming a neighbourhood by the wrong name can be an issue for some users

    maps:

    • representing maps for screenreader users remains an unsolved problem
    • before google maps became the #1, mapquest maps were big (before js; all HTML and images); in parts had better a11y features than google maps has today

    code of ethics in journalism

    • today's journalists aim to produce journalism that is accessible (it hasn't always been in the past)

    always important to remember: accessibility is not binary

    • there are always compromises
    • devs don't touch the a11y stuff because they are afraid of doing it wrong (wrong belief that a11y is either done right or bad, not "as good as possible")

    conclusion:

    • potentially a bias in the data journalism scene; it is easy to oversee that users may have limitations in cognitive or physical capabilities, compared to journalists and hackers
    • however, great awareness and interest for the issue (probably linked to ethics and self-understanding of journalists)
    • willingness to address the issue, but lack of (access to) useful resources how to do accessible data visualisation (as our discussion has shown, there are a lot of unsolved questions)
    • plans for a hacks&hackers berlin meetup in 2019 to discuss these issues; currently doing background research and looking for (berlin-based) speakers
    • overall agreement that this is an important (and exciting) topic for further discussion and exploration