settings can lead to unethical behavior; and how
properties that are external to the content, such as its
presentation or recency, effect the perceived credibility of communicated information or affect its impact
on other individuals.
The second half of the workshop began with a
keynote by Christo Wilson (Northeastern University),
who talked about methodologies for auditing biases
in social systems and algorithms. Participants then
discussed potential ways to train algorithms that are
more robust to perceptual biases present in data and
provide tools that would better inform end users. For
example, a corrective mechanism was proposed to de-bias data before training machine-learning models on
it. Other examples include tools like Buzzfeed’s “
outside your bubble,” meant to provide people with the
perspective of others, outside their siloed conversation on social media.
We intend to follow up on these ideas with participants and pursue a publication in the form of an
opinion piece, summarizing current knowledge and
possible future research direction for the study of perceptual biases and social media.
Nir Grinberg, Kenneth Joseph, and Brooke Foucault Welles organized this workshop and wrote this
report.
Social Media and
Demographic Research
Demography has been a data-driven discipline since
its birth. Data collection and the development of for-
mal methods have sustained most of the major
advances in our understanding of population process-
es. The global spread of social media has generated
new opportunities for demographic research, as indi-
viduals leave an increasing quantity of traces online
that can be aggregated and mined for population
research. At the same time, the use of social media
and the Internet are affecting people’s daily activities
as well as life planning, with implications for demo-
graphic behavior. The goal of this workshop was to
favor communication and exchange between the
communities of demographers and data scientists. It
revolved around the main theme of applications and
implications of social media and online data for
demographic research.
The workshop was organized by Emilio Zagheni
(University of Washington, Seattle), Ingmar Weber
(Qatar Computing Research Institute, HBKU), and
Thomas LeGrand (Montréal University, Canada). No
report was submitted by the organizers.
Studying User Perceptions
and Experiences with Algorithms
From Facebook’s News Feed algorithm that shapes the
posts and updates we see, to Spotify’s recommenda-
tion service that introduces us to new music that we
might love, to dating site algorithms that attempt to
match us with potential romantic partners, algo-
rithms play an increasingly important role in shaping
many aspects of our daily lives. The Studying User
Perceptions and Experiences with Algorithms work-
shop brought together a community of researchers
interested in taking a human-centered perspective on
studying the experience of algorithms.
The objective of this workshop was to articulate the
grand challenges of studying the user-algorithm relationship and to bring together participants interested
in developing projects to address these grand challenges. During the first breakout session of the workshop, participants identified a number of outstanding
research questions in this area. These included questions such as: What do users think is an algorithm?
How do users employ their (mis)understandings of
how algorithms work to reverse engineer or manipulate them? And does new information about how to
manipulate algorithms change users’ perceptions of
how the algorithm works? How do values and preferences transfer from people to algorithms? How do different degrees of awareness of algorithms change user
behavior? How do algorithms obscure themselves?
What makes users hostile or positively disposed to an
algorithm? What parts of an algorithm “should” users
see or not see? And who should be in charge of making these decisions? How can we combine “big data”
methods with “small data” methods to discern
longer-term effects of information filter algorithms
on users’ worldviews?
Participants then self-selected groups for the second breakout session based on the tractable entry
points they found to be of interest during the first
breakout. Groups used this time to incubate research
ideas, focusing on how they might take action on the
questions. At the end of the session, groups reported
out on the projects they envisioned.
Participants were encouraged to develop a short
abstract for the project on user-algorithmic interaction they envisioned, using Heilmeier’s questions
( www.darpa.mil/work-with-us/heilmeier-catechism)
as a guide. In addition to reporting back to the larger
workshop about their envisioned projects, participants also developed a short recommended reading
list ( www.studyingusers.org/reading-list) of articles
on this subject matter.
The cochairs for this event were Nicholas Proferes
(University of Maryland), Alissa Centivany (Western
University), Caitlin Lustig (University of California
Irvine), and Jed Brubaker (University of Colorado
Boulder). Additional organizers included Lala Haji-bayova (Kent State University), Marina Kogan (
University of Colorado Boulder), Tanushree Mitra (
Georgia Institute of Technology), and Nicole Ellison
(University of Michigan). This report was written by
Nicholas Proferes.