initial import into git of first draft

[bmh-research_statement] / research_statement.tex

\documentclass[10pt]{memoir}

% based on kieran healy's memoir modifications
\usepackage{mako-mem}
\chapterstyle{article-3}
\pagestyle{memo}

\usepackage{ucs}
\usepackage[utf8x]{inputenc}

\usepackage[T1]{fontenc}
\usepackage{textcomp}
\usepackage[garamond]{mathdesign}

\usepackage[letterpaper,left=1.2in,right=1.2in,top=1.2in,bottom=1.2in]{geometry}

% packages i use in essentially every document
\usepackage{graphicx}
\usepackage{wrapfig}
\usepackage{enumerate}

% packages i use in many documents but leave off by default
% \usepackage{amsmath, amsthm, amssymb}
% \usepackage{dcolumn}
% \usepackage{endfloat}

% import and customize urls
\usepackage[usenames,dvipsnames]{color}
\usepackage[breaklinks]{hyperref}

\hypersetup{colorlinks=true, linkcolor=Black, citecolor=Black, filecolor=Blue,
    urlcolor=Blue, unicode=true}

% add bibliographic stuff 
% \usepackage[round]{natbib}
% \def\citepos#1{\citeauthor{#1}'s (\citeyear{#1})}
% \def\citespos#1{\citeauthor{#1}' (\citeyear{#1})}

% import vc stuff after running `make vc`: \input{vc} \pagestyle{kjhgit}

\begin{document}

\setlength{\parskip}{4.5pt}

\baselineskip 14.5pt

\title{Research Statement}
\author{Benjamin Mako Hill}

\maketitle

My research is focused on collective action in online communities and
seeks to understand why some attempts at collaborative production --
like Wikipedia and Linux -- build large volunteer communities while
the vast majority never attract even a second contributor. I am
particularly interested in how the design of communication and
information technologies shape social outcomes like the decision to
join a community or contribute to a public good. My research is deeply
interdisciplinary and lies at the intersection of sociology,
communication, and human-computer interaction. I combine data from
online communities that make failures of collective action newly
visible with ``big data'' research methods from software engineering
to answer fundamental social scientific questions.

Seeking to understand the determinants of collective action, my
research has been driven by three overlapping themes: (1)
population-level observational studies comparing failures to build
communities to the rare successful attempts; (2) attention to the role
of reputation and status in collective action; and (3) analyses of
design changes treated as ``natural experiments'' to build a deeper,
and often causal, understanding of social processes from observational
data. Almost all of my work incorporates at least two of these themes.

\section{Population-Level Observational Studies}

Although there have been many thousands of studies of online
collective action and peer production, the vast majority of these
studies have only considered successful projects like Wikipedia and
GNU/Linux.  The majority of research on collective action -- both
online and off -- has only considered projects that have successfully
mobilized contributors. In this sense, most previous analyses of
collection action have systematically selected on their dependent
variable. Most of my research treats projects as the unit of analysis
and mobilization as the dependent variable, comparing successful
examples of collective actions to failures.

% \begin{wrapfigure}{r}{0.4\textwidth}
%  \begin{centering}
%  \includegraphics[width=2.4in]{figures/wp_citations_by_year.png}
%  \caption{Number of published academic articles with ``wikipedia''
%  in title by year.}
%  \label{fig:wppapers}
%  \end{centering}
%\end{wrapfigure}

\begin{wrapfigure}{r}{2.6in}
 \begin{centering}
 \includegraphics[width=2.6in]{figures/scratch_screenshot_default.png}
 \caption{A screenshot of the Scratch desktop programming environment
   where users create animations and interactive games.}
 \label{fig:scratchapp}
 \end{centering}
 \vspace{-2em}
\end{wrapfigure}


For example, in a working paper that is part of my dissertation, I
compare Wikipedia to seven attempts to create online collaborative
encyclopedia projects that were launched before Wikipedia
\cite{hill_almost_2012}. Using a inductive, grounded-theory based
analysis of founder interviews and archival data, I propose four
hypothesis for why Wikipedia built a large community while extremely
similar projects struggled to do so. Although the paper's methods
diverge from the quantitative, ``big data'' approach typical of most
of my work, the research question and strategy is representative.

I have also addressed this question in a series of quantitative
studies of the Scratch online community: a public website with a large
community of users who create, share, and remix interactive media. The
community is built around the Scratch programming environment: a
freely downloadable desktop application that allows amateur creators
to combine media with programming code (see Figure
\ref{fig:scratchapp}. Despite the fact that Scratch is a community
designed to promote collaboration through content remixing, only about
ten percent of Scratch projects will attract even a second
contributor.

\begin{wrapfigure}{l}{2.6in}
 \begin{centering}
 \includegraphics[width=2.6in]{figures/frontpage_modified-topremix.png}
  \caption{The front page of the Scratch online community where users
    can share and collaborate on projects.}
 \label{fig:scratchfrontpage}
 \end{centering}
 \vspace{-2em}
\end{wrapfigure}

In one study, forthcoming in American Behavioral Scientist, I test
three of the most commonly cited theories associated with
``generativity'' (i.e., qualities of technology or content that make
some works more fertile group for collaboration). I find some support
for previous theories but also find that, across the board, factors
associated with increased collaboration are associated with less
original and less transformative modes of joint-work
\cite{hill_remixing_2012}. In another published study using data from
Scratch, I show that more superficial collaboration leads to negative
reactions and displeasure in the community
\cite{hill_responses_2010}.

I am conducting a similar population based analysis in a new dataset I
have created with a population of 80,000 public attempts at ``wikis''
(i.e., public websites using the same infrastructure as Wikipedia). In
a working paper using this dataset, I consider inter-organizational
effects of competition for volunteer labor and find little support for
a widely cited ecological model of collective action from sociology
that treats volunteer labor as fixed and finite resource. Using data
on wikis hosted by both Wikia and the Wikimedia Foundation, I show
that contributions across peer production projects are driven by
environment-level changes in interest and that projects can even
benefit from complimentarities between similar projects
\cite{hill_is_2012}.

\section{Reputation and Status}

Although empirical research comparing successful and unsuccessful peer
projects has been rare, theories offering to explain the success of
collective action online have been widespread. No theory has been more
influential or widespread than the suggestion that, in the absence of
pecuniary rewards, engagement in collective action in online
communities is driven by the possibility of increased reputation and
status conferred upon contributors to an online public good.

\begin{wrapfigure}{r}{0.3\textwidth}
 \vspace{-1em}
 \begin{centering}
 \includegraphics[width=1.9in]{figures/barnstar_alone.png}
 \caption{Image of a ``barnstar'' social award given by Wikipedia
   contributors to each other to recognize positive contributions .}
 \label{fig:barnstar}
 \end{centering}
 \vspace{-1em}
\end{wrapfigure}

In a study of status-based awards in Wikipedia called ``barnstars''
(see Figure \ref{fig:barnstar}) that I will be submitting to a major
sociology journal by the end of this year, I provide an empirical test
of an influential status-based theory of collective action from
sociology. Although the study finds support for the widely
hypothesized virtuous cycle of status rewarding active contributors,
it also finds that these effects are limited to a sub-population of
Wikipedia contributors -- i.e., those who show their awards off to
others in virtual trophy cases \cite{hill_status_2012}. This result
has broad implications for both status-based theories of collective
action as well the design of reputation-based reward systems.

In another study of Scratch, nominated for best paper at the CHI 2011
conference \cite{monroy-hernandez_computers_2011}, I paired a
quantitative analysis of a design change with in-depth interviews of
users to demonstrate how credit-giving is ineffective when it stems
from an automated system because the system fails to reinforce
status-ordering through credible human expressions of social deference
and gratitude.

\section{Design-Driven Natural Experiments}

\begin{wrapfigure}{r}{0.25\textwidth}
 \begin{centering}
 \includegraphics[width=1.5in]{figures/lilypad.png}
 \caption{A image of the LilyPad Arduino microcontroller.}
 \label{fig:lilypad}
 \end{centering}
\end{wrapfigure}

Although nearly all of my work has important implications for the
design of socio-technical systems, I have structured much of my work
around the evaluation of design interventions. In several papers, I
treat design changes in technologies as ``natural experiments'' that
exogenously change the ways that social structure is enacted in order
to both build causal understanding from field data and to tighten the
distance between theories and and design.

For example, to evaluate the impact of status-based incentives and
collaboration in Scratch, I use a regression discontinuity framework
to measure the causal effect of increased status for collaboration in
Scratch \cite{hill_causal_2012}. In that study, which I am preparing
for submission to a communication journal this fall, I show that the
highlighting the work of the authors of collaborative projects on the
Scratch web page (see the bottom of Figure \ref{fig:scratchfrontpage})
resulted in more collaboration but also caused a decrease in the
amount of total effort exerted by contributors. Speaking to
fundamental sociological work in the literature on collective action,
I present evidence that this decrease is driven by both an the influx
of new contributors and a decrease in the effort and contributions of
established contributors.

In other papers, I have helped analyze sales records of hobbyists
microcontrollers to suggest that relatively simple design changes in
the \emph{LilyPad Arduino} -- a electronics toolkit minimally
re-designed for women and girls (see Figure \ref{fig:lilypad}) -- can
lead to large increases in the proportion of women contributors
through important shifts in the type of projects created
\cite{buechley_lilypad_2010}. In other work, I have explored how
technical errors may be able to provide similar opportunities for
analysis \cite{hill_revealing_2010}.

% or changes in socio-technical systems describing responsibility for a piece of software can lead to an important impact in the type and structure of contributions in peer production \cite{michlmayr_quality_2003}

\section{Research Agenda}

My research agenda involves further exploration of the determinants of
collection action online -- especially using a series of large new
datasets I have built in my dissertation research. I plan to both
continue this research trajectory and to create new social and
technical infrastructure that will allow others researchers to join me
in the type ``big data'' observational research in active communities
that my research involves. This section outlines some future
directions I plan to explore.

\emph{Toolkits for Experimental Social Design} -- My research has been
possible through relationships I have with a series of organizations
with large active online communities (e.g., the MIT Media Lab and the
Wikimedia Foundation). These organizations, like many others, make
design changes to the software that supports their communities to
encourage contributions and improve aspects of their users'
experiences. Most of the time, these organizations have very little
idea if these changes work. I plan to build on my own experience
evaluating natural experiments in online communities to create a
technical framework, and a network of academic and practitioner
collaborators, to facilitate well-designed experiments by the hosts of
large online communities and a system for data sharing that allows for
academic evaluation of these experiments.

\emph{Understanding the Relationship Between Collective Action and
  Performance} -- My work has treated collective action and production
as ends in themselves and has largely avoided the consideration of
issues of performance, efficiency, or quality. Using my existing
datasets, I will compare the performance of collaborative production
to individually produced works to understand when successful
collection action leads to higher performance or quality. In a
manuscript currently under review using data from Scratch, I show
important limitations of collaboration in remixing quality,
particularly in regards to more artistic or media-intensive projects
\cite{hill_cost_2012}. I will further explore this direction in future
work.

\emph{Integrated Theory of Design for Collective Action} --
My studies of status provide a detailed understanding of the dynamics
of collection action in relation to one important independent
variable. In future work, I plan to evaluate the effect of governance
and different systems of authority, framing, modularity and project
complexity. In the long term, I hope to work toward a broad set of
principles of design for collection action and community.

In graduate school, I have been fortunate to collaborate with many
co-authors in many academic departments and I intend to continue this
tradition going forward. In sum, my research uses design to contribute
to social scientific theories of collective action, and using theories
of collective action to influence design and offers implications and
opportunities for a broad range of disciplines and fields.

% bibliography here
\renewcommand{\bibsection}{\section{\bibname}\prebibhook}
\baselineskip 14.2pt
\bibliography{refs-processed}
\bibliographystyle{unsrt}

\end{document}