Understanding and Improving Social Interactions in Online Participation Platforms
Understanding and Improving Social Interactions in Online Participation Platforms
Understanding and Improving Social Interactions in Online Participation Platforms
The research scope of this project lies at the intersection of two different areas: computational social science and machine learning. The main goal is to develop the required tools to analyze, model and influence social behavior in order to optimize performance of a given online platform. The project will focus in online platforms for open democracy, but other types of platforms such as news aggregators, Wikipedia or MOOCs will also be analyzed.
The platform considered in this project is open source and has been adopted by the area of citizen participation and transparency of Madrid as Decide Madrid and the City Council of Barcelona (Decidim Barcelona). Decide Madrid was launched in September 2015 and by Jan 2016 involved more than 400 debates, 6,066 proposals and more than 10,000 registered users. Decidim Barcelona was launched in Feb 2016 within the municipal actuation plan including more than 10,000 proposals.
The source code of the platform is open source and has been already forked more than 100 times. We expected a critical mass to be reached as the first proposals made their way to the parliament. Our role in both platforms is to suggest and implement possible features or extensions to improve its performance. In research terms, this means that the platform can be used, not only as a source of high-quality social data, but also, given our possibility of performing interventions, to conduct controlled experiments in which we analyze the influence of different factors in a social environment.
The datasets generated by these platforms are valuable in a broad scope because (a) they can be used for evaluation of methods / algorithms developed not only by the research groups at the UPF, but by many others and (b) they represent a unique source of real social interaction data in the form of discussions which has potential interest in sociology, politics and psychology.
Some of the research topics and questions that are being addressed by the project include (in brackets the involved DTIC groups):
-
development of machine learning tools to increase user participation and satisfaction with the platform, i.e. user profile modelling, recommender systems, preference modelling, etc.. (AI/SM/DM)
-
probabilistic modelling of consensus formation (AI/SM/DM)
-
learning patterns of undesired behavior in online conversations and developing tools to prevent it (AI/SM/DM)
-
understanding the mechanisms of competition between different online resources, in this case political proposals compete for the attention of the users (SM)
-
bandit algorithms for website optimization (AI)
-
development of effective measures for user reputation (SM/DM)
-
probabilistic modelling of exogenous influences using latent sparse features (AI/DM)
-
sentiment analysis of the individual comments and their predictive power (AI/TALN)
-
automatic summarization of discussions (AI/TALN).
Short term impact: A number of datasets will be created and analyzed by the involved research groups. The datasets will be composed of the digital traces of interaction occurring in the platform during a period of time. In addition to the datasets, a number of tools to visualize and monitor the discussions will be made available. Examples of such tools already developed are: threads, explorer, user network, tag network. The datasets will be released and made public.
The platform has been built within the existing EU project DCENT (Decentralised Citizens ENgagement Technologies) at EURECAT, which participates as a third party in the project. DCENT will finish in June 2016.
To know more:
-
Presentation of the project at the Data-driven Knowledge Extraction Workshop, June 2016 (Slides)
-
Github (datasets and tools): https://github.com/elaragon
-
Data exploration for Meneame dataset: http://mdm.barcelonamedia.org/meneame/
-
Eurecat web: http://eurecat.org/en/