Cooperative System Design between Citizens and Organizatons

As the Web of data is growing, especially unstructured social media data, it has presented an information overload challenge for answering a variety of questions for organizational sensemaking: what is actionable in data streams to help organizational workflow (for meeting organizational task 'articulation'), how to filter, whom to engage with in the social community (for enhancing 'awareness'), etc.

A citizen sensor community (CSC) presents data level challenges also: Ambiguity in interpreting unconstrained natural language text, sparsity of user behaviors, and diversity of user demographics. Interdisciplinary research involving social and computer sciences is essential to address these socio-technical issues.

We focus on a cooperative web information system design that helps in sensemaking of user-generated big data to manage dependencies of organizational coordination by mining articulated information needs from an informal CSC on Web 2.0 platforms. For example, during crisis response coordination such dependencies include: what (are resource needs/availability), where (are resources needed/available), who (needs/offers resources to help), and when, due to multiple actors involved for resource management from both communities—the informal social & the formal response. Uncoordinated engagement between these two communities can lead to a second disaster; for example, as NPR reported after Hurricane Sandy that piles of unsolicited, unnecessary clothes were lying for responders to manage.

We propose an Identify-Match-Engage computing framework to improve cooperation between the formal organizational and the informal social communities. Identifying intents related to dependencies on resources to meet 'articulation' of organizational needs, matching those intents, and engaging with the prioritized set of actors for enhancing 'awareness' in the informal social community enable greater cooperation.
The IME framework allows operationalizing computation in design problems of 'awareness' and 'articulation' of the cooperative system between citizens and organizations, by addressing data problems of group engagement modeling and intent mining.

• H. Purohit. Humanitarian Informatics Approach to Cooperation between Citizens and Organizational Decision Makers. CADMICS'16 in conjunction with ACM CSCW, 2016.

• H. Purohit, M. Dalal, P. Singh, B. Nissima, V. Moorthy, A. Vemuri, V. Krishnan, R. Khursheed, S. Balachandran, H. Kushwah, A. Rajgaria.Empowering Crisis Response-Led Citizen Communities: Lessons Learned from JKFloodRelief.org Initiative. In Strategic Management and Leadership for Systems Development in Virtual Spaces, IGI-Global, 2015.

• H. Purohit. Mining Behavior of Citizen Sensor Communities to Improve Cooperation with Organizational Actors. Ph.D. Dissertation, Dept. of Computer Science and Engineering, Wright State University, 2015.

• T. Clark, C. Kessler, H. Purohit. Feasibility of Information Interoperability in the Humanitarian Domain. AAAI 2015 Spring Symposium on Structured Data for Humanitarian Technologies.

• First AAAI Symposium on Structured Data for Humanitarian Technologies, #SD4HumTech15.

Behavior Mining from Unstructured, Informal Social Data

Mining behavior from user-generated text is challenging due to the high level of subjectivity involved, and also due to the fact that, by human nature, we often express goals with different degrees of clarity. Furthermore, the informal nature of social media data makes it more challenging.

We tackle the problems of mining intent at the message level, using novel methods to identify within-message-level actor-specific intents, as shown in the example above. Using psycholinguistic knowledge of intent expressing patterns, We combine the behavior knowledge with the machine learning and NLP approach to mine intents in the data.

Publications:

• H. Purohit, C. Castillo, F. Diaz, A. Sheth and P. Meier. Emergency-Relief Coordination on Social Media: Automatically Matching Resource Requests and Offers. First Monday journal, Vol. 19, Issue 1, Jan 2014. (Highlight of the FM January Issue)
• H. Purohit, A. Hampton, S. Bhatt, V. Shalin, A. Sheth, J. Flach. Identifying Seekers and Suppliers in Social Media Communities to Support Crisis Coordination. Journal of CSCW, Springer, 2014. (Selected in Special Issue for Crisis Collaboration)
• H. Purohit, G. Dong, V. Shalin, K. Thirunarayan, A. Sheth.Intent Classification of Short-Text on Social Media. The 8th IEEE International Conference on Social Computing and Networking, 2015.

Individual and Group Engagement in Social Media Communities

Social media communities formed around discussions of dynamic events present an interesting phenomenon of individual engagement. We explore the questions of 'who' will engage in the community, 'how much' an individual will engage in the community, and which groups will coherently engage over time in the social community discussions during real-world events.

INDIVIDUAL BEHAVIOR:

In the following work (Figure 2) on mining Twitter users' followers such that 'who' and 'how much' will engage in the discussion, we explored the interplay of people, content, and network features. We concluded that every type of feature cannot be expected to perform well in all types of events such as Global vs. Local, Transient vs. Lasting, etc.

• Purohit, H., Ruan, Y., Joshi, A., Parthasarathy, S., & Sheth, A. (2011). Understanding user-community engagement by multi-faceted features: A case study on twitter. In WWW 2011 Workshop on Social Media Engagement (SoME).
• Ruan, Y., Purohit, H., Joshi, A., Parthasarathy, S., & Sheth, A. (2012). Prediction of Topic Volume on Twitter. 4th Int'l ACM Conference on Web Science, WebSci-2012. (Acceptance Rate: 27%)

GROUP BEHAVIOR:

In the following work (Figure 3) on mining group engagement behavior over time, we analyzed which groups would coherently interact over time, and proposed a content-driven metric for group discussion divergence to model social dynamics, complementing structural metrics in the past work. We further asked the question, if there is an influence of group members' social identity and cohesion on this engagement pattern.

We observed that groups with increasing divergence over time post past episodic events related content, whereas the groups with decreasing divergence write more content with in the nature of future actions and social interactions.

• Purohit, H., Ruan, Y., Joshi, A., Parthasarathy, S., & Sheth, A. (2014). On Understanding Divergence of Online Social Group Discussion. In 8th AAAI Int'l Conference on Weblogs and Social Media, ICWSM-2014. (Acceptance Rate: 23%)

#SocialGood: Mining Society Beliefs about Gender-based Violence

With guidance from subject matter experts at the United Nations Population Fund (UNFPA), We are studying social big data (recent work consisting of 14 million tweets collected from the Twitter platform over 10 months), to analyze public opinions on the sensitive issue of gender-based violence (GBV). It highlights relevant key topics and the nature of tweeting practices by geography, time, and events related to various societal beliefs.

The objective is to glean insights about public awareness on certain sensitive topics (for example views on rape and the safety of women) and also inform data-driven insights to (non-)governmental organizations for assisting policy and law-making. It can also help devise preventive campaigns and strategies to improve public sensitivity and awareness toward harmful practices against women.

Among such insights, we found empirical evidence that people use sensitive topics as metaphor highest in the sports domain, which informs decisions regarding the creation of target campaigns during sports events to bring about public awareness and curbing gender-based violence.

• Social Media Content:
  • Substantial GBV related content exists in social media.
  • Spikes in GBV content reflect the influence of transient events, particularly involving celebrities.
  • Gender, language, technology penetration, and education influence participation with implications for the interpretation of quantitative measures.
  • GBV content includes humor and metaphor (e.g. in Sports) that reflect both attitude and behavior.
  • Content highlights the role of government, law enforcement and business in the tolerance of GBV.
• Relevance to GBV Policy:
  • Social Media provides an alternative for measuring GBV attitude and behavior that is cheaper, faster, and broader than conventional survey-based methods.
  • Regional socio-cultural context influences both the measurement and the interpretation of data.
  • Computational methods for context-sensitive monitoring, modeling and interpreting GBV social media content are highly feasible.
  • The location and network of social media participants supports targeted regional anti-GBV campaigns.
  • Policy makers require tools to make social media content accessible in near-real time to monitor the effectiveness of anti-GBV campaigns.

• Purohit, H., Banerjee, T., Hampton, A., Shalin, V. L., Bhandutia, N. & Sheth, A.P. (2016). Gender-Based Violence in 140 Characters or Fewer: A #BigData Case Study of Twitter. First Monday, 21 (1).