To Call, or To Tweet? Understanding 3-1-1 Citizen Complaint Behaviors

phone services offer citizens a crowdsourced platform to call and report about issues in their communi- ties. In recent years, local governmentsand agencies have started to offer new communication channels through social media including Twitter accounts. Although not its main purpose, citizens are using these channels as a way to communicate 3-1-1 service requests to their local authorities. Furthermore, Twitter is also being used by citizens to share issues about their communities with friends and colleagues, without specifically addressing it to any local authority. In this paper, we analyze the behavioral similarities and differences between the use of formal 3-1-1 phone services and informal channels - like Twitter - to report about issues that affect a community. Our end objective is to help public institutions understand the relevance of informal communicationchannelsas data sources for service requests. For that purpose, we design and evaluate a set of supervised classifiers that automat- ically label tweets as complaints and determine its type. A weighted multiclass classifier was selected based on its performance with precision and recall values of 86% and 62%, respectively. By comparing labeled tweets against official 3-1-1 phone service request records, we provide a large-scale analysis of citizen complaint behaviors over the two crowdsourced channels. I INTRODUCTION In 1996, the city of Baltimore (Maryland) was the first to deploy a 3-1-1 special phone number for non-emergency service requests. Citizens could call to report situations or complaints regarding a wide range of issues from traf- fic to noise or heating problems in their buildings. This initial deployment was run by the local police department and had an impressive initial success (22). Since then, similar services have sprung out through the US and in many other countries, sometimes under different num- bers. The most important contribution of 3-1-1 phone services is that they constitute a single point of contact to the different agencies that handle each type of service complaint. Instead of having citizens memorize individ- ual agencies, they can just call and report a complaint that will be addressed in a transparent manner by the cor- responding agency. For example, 3-1-1 can contact the Department of Transportation to address issues regarding real-time traffic, ferries or biking; similarly, they can con- tact the Taxi and Limousine Commission to solve issues that relate to taxis and VIP transportation services. But above all, 3-1-1 phone services share an important phi- losophy: to give tools to citizens to report about issues in their communities. As such, 3-1-1 phone services consti- tute an early form of crowdsourcing with citizens report- ing non-emergency situations that matter at a local level. With new times come new services. Local governments and agencies have adapted to the times and now offer new communicationchannels through, for example, Twit- ter accounts. These channels, which were originally in- tended to disseminate information and reach citizens, are not scrutinized in real-time and 24/7 by local agencies. However, citizens are increasingly using these channels as crowdsourced tools to report 3-1-1 service requests to agenciesandlocal governments. Furthermore,individuals are also using their own social media accounts to report issues in their communities and share them with their fol- lowers and friends, often times without explicitly address- ing them to an agency'saccount. As a consequence, given the popularity of social media and the increasing avail- ability of smartphones, it becomes critical for agencies and local governments to understand how social media channels like Twitter are being used to communicate 3-1- 1 complaintseither directly to the agenciesor indirectlyto followers and friends. In fact, such analysis could prove determinant to evaluate whether city halls should follow and give service to social media channels in real-time, ei- ther with humans (like the 3-1-1 phone) or automatically through data analytics. In the future, fully responding to social media service requests might be the only way for local institutions to manage the millions of potential calls they would get without these alternative communication channels put in place. In this paper, we focus our analysis on understanding the behavioral similarities and differences between the use of 3-1-1 phone services and Twitter channels to report is- sues that affect a community. The former constitutes the formal (official) channel to report complaints or service requests while the latter, although periodically screened, represents an informal (unofficial) channel. Our research

[1]  Eriq Augustine,et al.  Outage detection via real-time social stream analysis: leveraging the power of online complaints , 2012, WWW.

[2]  Jianhua Lin,et al.  Divergence measures based on the Shannon entropy , 1991, IEEE Trans. Inf. Theory.

[3]  McElory Hoffmann,et al.  Unsupervised Construction of Topic-Based Twitter Lists , 2012, 2012 International Conference on Privacy, Security, Risk and Trust and 2012 International Confernece on Social Computing.

[4]  JapkowiczNathalie,et al.  The class imbalance problem: A systematic study , 2002 .

[5]  Susan T. Dumais,et al.  Characterizing Microblogs with Topic Models , 2010, ICWSM.

[6]  Sean A. Munson,et al.  Social Media Technology and Government Transparency , 2010, Computer.

[7]  Kyumin Lee,et al.  You are where you tweet: a content-based approach to geo-locating twitter users , 2010, CIKM.

[8]  John Platt,et al.  Probabilistic Outputs for Support vector Machines and Comparisons to Regularized Likelihood Methods , 1999 .

[9]  Hila Becker,et al.  Hip and trendy: Characterizing emerging trends on Twitter , 2011, J. Assoc. Inf. Sci. Technol..

[10]  Yulan He,et al.  Online Sentiment and Topic Dynamics Tracking over the Streaming Data , 2012, 2012 International Conference on Privacy, Security, Risk and Trust and 2012 International Confernece on Social Computing.

[11]  Mark Dredze,et al.  You Are What You Tweet: Analyzing Twitter for Public Health , 2011, ICWSM.

[12]  Isabell M. Welpe,et al.  Predicting Elections with Twitter: What 140 Characters Reveal about Political Sentiment , 2010, ICWSM.

[13]  Nathalie Japkowicz,et al.  The class imbalance problem: A systematic study , 2002, Intell. Data Anal..

[14]  Tamara A. Small WHAT THE HASHTAG? , 2011 .

[15]  Alexander Schellong,et al.  Managing Citizen Relationships in Disasters: Hurricane Wilma, 311 and Miami-Dade County , 2007, 2007 40th Annual Hawaii International Conference on System Sciences (HICSS'07).

[16]  Ryan M. Rifkin,et al.  In Defense of One-Vs-All Classification , 2004, J. Mach. Learn. Res..

[17]  Stefan Savage,et al.  Measuring Online Service Availability Using Twitter , 2010, WOSN.

[18]  B. Loader,et al.  What the hashtag? A content analysis of Canadian politics on Twitter TAMARA A . SMALL , 2012 .

[19]  Peter Willett,et al.  Readings in information retrieval , 1997 .

[20]  Salvatore J. Stolfo,et al.  A Network Access Control Mechanism Based on Behavior Profiles , 2009, 2009 Annual Computer Security Applications Conference.

[21]  Yang Liu,et al.  "I loan because...": understanding motivations for pro-social lending , 2012, WSDM '12.

[22]  Andree Woodcock,et al.  VoiceYourView: anytime, anyplace, anywhere user participation. , 2012, Work.

[23]  Henry A. Kautz,et al.  Modeling Spread of Disease from Social Interactions , 2012, ICWSM.

[24]  Ana-Maria Popescu,et al.  A Machine Learning Approach to Twitter User Classification , 2011, ICWSM.