=Paper= {{Paper |id=Vol-1362/USEWOD_paper6 |storemode=property |title=None |pdfUrl=https://ceur-ws.org/Vol-1362/USEWOD_paper6.pdf |volume=Vol-1362 }} ==None== https://ceur-ws.org/Vol-1362/USEWOD_paper6.pdf 
How will we interact with the #WebWeWant?
Position statement for the 5th International USEWOD Workshop: Using the Web
in the Age of Data, May 31st, 2015, Portoroz, Slovenia.

Position paper


Power to the agents?! In the #WebWeWant, people
will critically engage with data – and data journalism
can help them want to do this
Bettina Berendt
Professor at the Department of Computer Science, KU Leuven

USEWOD’s premise has been, from the outset, that the Web of Data – just like the Web as
a whole – only makes sense to the extent that it is used. But who or what should do the
using? As Ruben Verborgh reminds us in his USEWOD 2015 blogpost, the original vision
was: “agents, which would use the Web to do things for people“. And as Ruben also points
out: as of now, there’s not much sign of such agents.

Or is there? Max van Kleek’s USEWOD 2015 blogpost recalls the strong agency that
platforms in the mobile-app domain have: as orchestrators, gatekeepers, and data
controllers. These platforms do things, and they do them for people: the providers of apps
and platforms, their owners, their shareholders, … Van Kleek warns of the consequences
that a repetition of this model would have when applied in the next big application domain
of networked services: the Digital Home. And he argues that the Web (the original one,
the human-readable one) would be a better model for this new domain. Thus, what’s
missing may not be “agents that use the Web to do things for people“, but “agents that
use the Web to do things for different people“ – not only the economically or politically
powerful, and also not only the technologically savvy.

To be fair, the vision of Semantic-Web software agents that Berners-Lee, Hendler and
Lassila developed in their 2001 Scientific American article on the Semantic Web (and that
Ruben refers to) is at least partially democratic in the sense that the authors talk about
“consumers agents“ and “producer agents“. They conceptualized the latter as being “the
agents of individual providers“, who supply information and transaction possibilities
“through their Web sites“. This decentralisation is of course representative of the vision of
Web and Semantic Web pioneers. It did not preview market concentration tendencies such
as those described by van Kleek, and could therefore not preview the power differentials
that we are observing today. One reason for this blind spot in expectations may be that
technocentric visions such as the 2001 Semantic-Web software agents overlook one basic
fact: The Web and all its relatives are complex socio-technical systems (see
Ramine Tinati’s USEWOD 2015 blogpost), in which software agents and human agents do
things and co-determine outcomes.

So how can we create a truly democratic Web in which agents do things “for all people“?
In a complex socio-technical system, there can be no easy answer to this question. But
whatever the answer(s), a new role for human users of the Web is essential:
someone who can critically examine not only verbal arguments as in the Web of
Documents, but also data, whether they are RDF LOD or structured data accessible
through some special-purpose API. This role is essential for debunking myths about data
being “objective“, “speaking for themselves“, and thereby giving us access to an
unequivocal truth. Instead, rather than naïvely assume such objectivity in data, we should
always critically question the origins of data, the purposes with which they were collected
and processed, and the methods with which this was done.
But how can we make people want to engage with data in the first place? We believe that
one way towards this complex and ambitious goal is via intermediaries who tell us
interesting, appealing, and ultimately thought- and criticism-provoking stories about data.
And we believe that today, data journalism is a premier type of such intermediaries,
because data journalists combine the expertise of how to access, interpret, and give
access to data, with the expertise of storytelling that can engage wide ranges of people.

By calling for data journalism in this comprehensive skill-set sense, are we not just shifting
the problem by entrusting a new elite of highly-educated people with explaining the (data)
world to us? A partial answer to this conundrum is the observation that data journalism is
usually produced in teams consisting of journalists, designers, and developers (with the
latter being the data experts in the computational sense). And this observation calls for
new forms of team education towards data competencies.

In the recent 2-day hackathon News Hack 2015, led by Kris Vanhemelryck, we have
tested our initial ideas for such an education. News Hack is part of the interdisciplinary
community Nieuwslab.be, in which seven degree programmes in the KU Leuven
Association collaborate via a blog, internships and workshops in order to stimulate
journalistic innovation.
News Hack 2015 assembled three interdisciplinary groups of 6-7 students each from
journalism, media studies, computer science, and related disciplines. Their task was to
“make an innovative, digital and transmedial news story starting from a top-level question
and initial datasets that you receive from the organisers. You choose the specific
questions, the form, perspective and tools that you use to tell this story.“ After three
hours of preparation, groups pitched their topics to the coaches and the plenum, and then
returned to their work and further research. They were able to use state-of-the-art
audiovisual tools and get the help from coaches throughout. The use of specific creativity
techniques and reporting methods were strongly encouraged. All groups employed live
and/or telephone interviews with citizens and other stakeholders of their stories. On the
afternoon of the second day, the final products were presented to a jury consisting of
journalists, scientists, and data scientists from industry. Prizes were awarded for best
story, best form, best research, best data use, and highest innovativeness.

The results were impressive . They show the potential of the “pressure cooker“ approach
of this intense two-day collaboration. A particularly interesting aspect are the different
ways in which the three teams used – or did not use – data to engage their readers
in storytelling:
The ‘Blijf van mijn lijf’ editorial team (“stay away from me”) reported on sexual
harassment, triggered by the #wijoverdrijvenniet (“we are not exaggerating”) trending
hashtag that started a few days before the hackathon, used by women to describe the
manifold experiences of everyday harassment they encounter. In addition to conducting
background research and interviews with experts, the team extracted 2,000 tweets and
processed them with natural-language techniques into a Dadaist poem. The poem is a
form of data visualisation that invites the reader to contemplate a random – and yet
representative?! or totally biased?! – sample from the continuing story of this hashtag that
keeps being told by Twitter users in their thousands (e.g., 18,000 tweets counted in a
week for #wijoverdrijvenniet or 57,000 for the earlier German equivalent #aufschrei;
more recent trending hashtags with similar topics include #dailyracism).
‘Kerk te koop’ (“Church for sale”) investigated possible futures for the 1800 Flemish
churches. Will they be remodelled as car repair shops, luxury hotels, or bookshops? The
team contacted a Church Real Estate Agent, interviewed people in the street, and gave
readers ways of interacting with data on church attendance in different life contexts over
the past 50 years.

The editorial team of ‘Reinvent Leuven’ investigated the telling of migrants’ stories. They
created a set of mini-documentaries, each guided by one person’s answer(s) to the
question “What do you miss in Leuven?”, and set up a Facebook Community.
Interestingly, this approach (inspired by the photoblog Humans of New York) uses no data
beyond some initial informational statistics, relying instead on the audiovisual potency of
the individual personal story for engaging its readers/contributors.

In sum, we believe that data journalism can be a vehicle for engaging more and more
diverse people into looking at, interacting with, and hopefully also questioning data. We
also believe that ideas and techniques of data journalism can be taught to interdisciplinary
teams in short and entertaining educational events such as the News Hack 2015
hackathon, and that such educational experiences can help shape the students’ further
attitudes to data in the news. (Events similar to News Hack 2015 have recently become
very popular, cf. Hackastory or the ONA Student Newsroom.)

Of course, much more work is needed to substantiate these claims. Expectations
regarding the extent to which media users actually do use, understand, or even question
data in data journalism must be tested empirically. Alongside such evaluations, tried-and-
tested approaches to good storytelling with data need to be assembled into a body of
knowledge, skills and tools that further creators can work with. As one contribution
towards this goal, we have started assembling and commenting on tools that we used in
and around News Hack 2015.
Finally, data journalism is only one approach to making people engage with data, and we
believe that the topic should ultimately find its way into more general curricula for digital
literacy (see here and here for teaching examples). For computer science, the question
will be how to develop next-generation software agents that meet the collaborative needs
of the newly empowered human agents.