Online Social Media

Digital Fingerprints of Cognitive Reflection

Co-follower Network of accounts that are followed by subjects in our study. The intensity of color of each node represents the average CRT of its followers (darker=higher CRT). Nodes are positioned using directed-force layout.
  • M. Mosleh, G. Pennycook, and D.G. Rand “Digital Fingerprints of Cognitive Reflection:‌ Cognitive Style and Content Shared on Twitter’’.
  • M. Mosleh, G. Pennycook, and D.G. Rand “Self-reported Willingness to Share Political News Articles in Online Surveys Correlates with Actual Sharing on Twitter’’ PsyArXiv.
  • Ziv Epistein*, M. Mosleh*, A. Arechar G. Pennycook, and D.G. Rand “Primed for Accuracy: A Cognitive Approach to Fighting Misinformation on Twitter’’.

Experimental Social Networks

Information Gerrymandering and Voting Coordination

We devised an online environment that enhances simultaneous interaction between individuals. We recruit players from online labour market (e.g., Mechanical Turk) and assign them to nodes on a social network through which they interact and receive information from other players and/or programmed AI players. Players receive different payoffs based on the collective outcome of the game. We developed a mathematical model to predict the outcome of the game for different parameters (incentive ratio, assortment within the social network,…). We use the data from experiments to update the mathematical model in an iterative process.

  • A. J. Stewart, M. Mosleh, M. Diakonova, A. Arechar, D. G. Rand, J. B. Plotkin “Information Gerrymandering and Undemocratic Decisions”.

Computational Social Networks

Co-evolution of Cognition and Social Norms on Structured Population

We interact and learn from our friends and they do the same with their own friends. Hence, the structure of social network play a key rule on the way we learn and adopt new behavior. We devised a multi-layered model that combines local agent interactions with social learning, thus enables both strategic behavior as well as diffusion of successful strategies within social network.

Natural Language Processing

Consumers Preferences in Airbnb vs. Tripadviser

Peer-to-peer platforms, collectively known as Sharing Economy, are becoming an inventible part of our daily economic activities and, in many cases, are substituting our usage of traditionally provided services by well-established firms. We investigate users’ opinion trends in a popular sharing economy platform for short term property rental (Airbnb) and compare it with a traditional platform for hotel reservation (TripAdvisor). We take a text analytics approach, extract opinion trends from online reviews, and compare them across different aspects of the service. We show that co-variation of opinion trends is a function of location and different aspects, which suggest a new way of market segmentation.

  • M. Mosleh and B. Heydari, “Co-evolution of Consumers Preferences in Sharing Economy and Traditional Platforms: the Case of Hospitality Industry” 3rd Annual International Conference on Computational Social Science (IC2S2). 2017, Cologne, Germany.
  •  M. Mosleh and B. Heydari, “Sharing Economy vs. Conventional Platforms: A Natural Language Processing Approach ” Workshops on Natural Language Processing and Computational Social Science at Annual Meeting of the Association for Computational Linguistics (ACL) 2017, Vancouver, Canada.

Economic Networks

Efficient and Stable Network Structures

There are many social,  economic, and technical situations where connection between two nodes in the network incurs a cost, yet facilitates overall resource access through out a network and comes with a benefit. Building upon economic network models, we found analytical solution for efficient and stable network structures and a wide range of cost and benefit functions.

Computational Models for Networked Systems Design

Distributed and networked architecture increases system flexibility and its responsiveness to environment uncertainties. Yet, an important question is how much flexibility is required given an uncertainty profile of the environment. We have developed an architecture framework that guides decisions about the level of modularity and distribution of networked systems and demonstrated it’s applicability to real world complex systems.

Data-driven Policy Analysis

Spatial Diffusion of Behavioral risk

By considering certain types of risk as social construct that emerge through social interactions over time, we study spatial diffusion of risky behavior in various geographic regions to study how changes in risky behavior in a region affect other regions. As the first step, we study diffusion of risky driving behavior among teenage drivers in the time period between 2000-2014 in the US.

Post Obama-Care Insurance Competition

One of the main goals of affordable-care act was triggering more competition among health insurance providers. We study the evolution of competition in health insurance market after introduction of ACA by studying the structural evolution of the network that connects participating insurance companies based on the similarity of their plans in similar regions. We demonstrate that this using network measures is a good predictor of evolution of competition in such markets.

  • D. Gianetto, M. Mosleh, B. Heydari, “Dynamic Structure of Competition Network in Affordable Care Act Insurance Market”, IEEE Access, 2018.