Concepts of collective intelligence (aka the "wisdom of crowds") play increasingly fundamental roles in understanding and modelling complex reasoning and evaluation scenarios, such as estimating the future value of investments, the replicability of scientific findings, outcomes of political races, how climate relates to immigration, and which cultural products will garner broad appeal. Understanding such collective reasoning scenarios is difficult, because they often violate two conditions that have been regarded by classic collective intelligence theories as necessary for the emergence of group evaluations that are superior to individual judgment: the diversity of information among group members and their independence when forming opinions. Within this context, my National Science Foundation-supported research addresses the puzzling open question: Under which conditions does collective intelligence provide accurate decision-making support for complex problems and when does it fail? In particular, I am investigating the impact of network embeddedness and diversity on collective intelligence, and devising ways to support collective reasoning in three high-stake real-world challenges: (1) scholarly communication, (2) online communities, and (3) cultural innovation.
My work is applying and developing quantitative network analysis techniques. I use theory-driven investigations that rely on mining Big Data about interactions and behavior, online experiments and surveys, novel statistical modelling frameworks (combining network science with, e.g., machine learning techniques), crowdsourcing, and text analysis. The core contributions of my work are relevant for Computational Social Science, Network Science, Social Computing, and Computer-Mediated Communication.
TRANSFORMING ONLINE SCHOLARLY COMMUNICATION
We investigate how scholars use social media in combination with online repositories to disseminate their publications to the general public and research communities. The proliferation of online platforms means that access to high quality scientific papers is potentially enhanced, even as the challenges for users to choose reliable and trustworthy sites intensifies. While scholars have readily adopted online methods of disseminating their work, there are no clear guidelines for how to do so in ways that ensure that the highest quality science receives the most attention. As the general public, journalists, and interested government increasingly turn to online sources of information, it becomes ever more critical that there be effective tools to help detect irresponsible research and publications. Within this context, we develop methods that investigate how scientific publications spread on various types of online platforms.
>> I Zakhlebin and E-Á Horvát, Diffusion of scientific articles across online platforms, AAAI Conference on Web and Social Media, Atlanta, GA, 2020
HARNESSING COLLECTIVE INTELLIGENCE IN ONLINE COMMUNITIES
Our work investigates empirically and at scale both successful and failed crowdfunding campaigns with the goal of exploring the consequences of factors that have been assumed to deter collective intelligence (i.e., low diversity and interdependent decisions). Based on longitudinal data from the oldest crowdfunding platform in the U.S., we showed that the wisdom of the lending crowd on the platform can provide reliable decision support with respect to which borrowers will default. Our findings indicate that a proper computational decoding and aggregation of informally communicated signals (e.g., trust in the campaign as indicated through the amounts invested and the times required to contribute those funds, lenders' tendency for herding) are more predictive of default than the available credit score, yet the latter is the one used broadly as an industry standard. Honing in on the role of social influence and networks, we demonstrated that social learning is essential for novice investors, while serial investors profit from a diversified portfolio. This research further supports the idea that norms and consensus have increasingly crystallized on crowdfunding platforms and indicates that social influence would unfold differently in the case of novel vs conventional projects. Our follow-up project investigated the latter possibility and found that novelty is evaluated differently by investors, but that overall it reduces significantly the chances of successful fundraising.
>> H Dambanemuya and E-Á Horvát, Harnessing collective intelligence in P2P lending, ACM Web Science, Boston, 2019 [pdf]
>> I Zakhlebin and E-Á Horvát, Investor retention in equity crowdfunding, ACM Web Science, Boston, 2019 [pdf]
>> E-Á Horvát, J Wachs, R Wang, A Hannák, The role of novelty in securing investors for equity crowdfunding campaigns, AAAI Conference on Human Computation and Crowdsourcing, Zurich, Switzerland, 2018 [pdf]
HOW NETWORK SCIENCE CAN HELP PROMOTE CULTURAL INNOVATION
I have an enduring passion for the organizing principles behind creativity and cultural industries. Across all of my theoretically and methodologically motivated projects on processes of collective evaluation, I thus seek connections and applications in creative sectors. Creative industries employ 4.9 million workers across the U.S. and have contributed $763.6 billion to the country’s economy in 2018, more than agriculture, transportation, or warehousing. Their societal role is key in shaping personal and collective identities, facilitating social inclusion, contributing to mental health, and supporting the economic regeneration of cities and regions. Yet, current cultural production often fails to fulfill its mission of animating communities, celebrating diversity, and improving quality of life. Properly evaluating cultural trends and understanding innovation opportunities has thus become critical for scholarship and leaders in creative occupations. My work in this space has looked at innovation through (1) influences from prior artworks and (2) collaborations between artists. Theoretically, these studies contribute to long-standing debates in the literature about an artwork's relation to its circumstances and the value of social capital in creative enterprises, advancing our understanding of the role of network embeddedness both at the level of artworks and artists. Ingrained in scholarship that demonstrates how creativity and innovation is facilitated by the recombination of previous ideas, my work in this space pushes the boundaries of what has been considered possible in terms of the scale and detail of the empirical validation of theories about cultural production.
>> Y Wang and E-Á Horvát, Gender differences in the global music industry: Evidence from MusicBrainz and The Echo Nest, AAAI Conference on Web and Social Media, Munich, Germany, 2019 [pdf]
>> A Spitz and E-Á Horvát, Measuring long-term impact based on network centrality: unraveling cinematic citations, PLOS One 9(10): e108857, 2014 [pdf]
>> A Spitz and E-Á Horvát, A cookbook of cinematic delicacies that do not expire, Leonardo 47(3):271, 2014 [pdf]