Periodic Reporting for period 1 - Attitude-Maps-4-All (Delivering a useable method for mapping groups in attitude networks by developing a robust prototype, development community, and exploitation model to maximize the social value of the breakthrough)
Período documentado: 2022-09-01 hasta 2024-02-29
Understanding group-based coordination of attitudes is critically important to our most pressing global issues including climate change, vaccine hesitancy and threats to democratic processes. Until now, it has been difficult to identify opinion-based groups and to link the micro to the macro (ie. locate individuals in group-based opinion systems), as methods tend to focus on one or the other level of analysis. For example, which attitudes are most important to group identity? How are individuals located within a group depending on their individual attitudes?
A core idea of the ERC-DAFINET project is that people are linked by the attitudes they jointly hold, and that attitudes become socially connected when they are jointly held by people. From this basic idea, we developed methods to construct bipartite networks directly from survey-based data to visualize attitude networks. This method has broad interdisciplinary relevance for visualizing group structure in survey data. For example, we have applied this method to attitude data collected in the first wave of the COVID pandemic in the UK (Maher, P. J., MacCarron, P., & Quayle, M. (2020). Mapping public health responses with attitude networks..... British Journal of Social Psychology, 59(3), 641–652. https://doi.org/10.1111/bjso.12396) and American political attitudes (Dinkelberg, A., O’Reilly, C., MacCarron, P., Maher, P. J., & Quayle, M. (2021). Multidimensional polarization dynamics in US election data..... Analyses of Social Issues and Public Policy, 21(1), 284–311. https://doi.org/10.1111/asap.12278). After deriving a graph from the survey data, we can apply network methods to detect polarization (even without extremism); identify items which most clearly distinguish groups; and track polarization over time. How much overlap is there between groups? How can we identify polarization across multiple opinion dimensions? The ERC-DAFINET project provides a breakthrough mathematical method to answer these questions, but previously required high levels of mathematical and coding expertise to implement from first-principles.
The Attitude-Maps-4-All Proof of Concept project prototyped open software for implementing this network method, making it available to ordinary end-users (researchers and practitioners across disciplines). Where it previously required high-level technical mathematical and coding skills to implement, as a result of the work done in this Proof of Concept grant, it is now accessible to any researcher.
A core idea of the ERC-DAFINET project is that people are linked by the attitudes they jointly hold, and that attitudes become socially connected when they are jointly held by people. From this basic idea, we developed methods to construct bipartite networks directly from survey-based data to visualize attitude networks. This method has broad interdisciplinary relevance for visualizing group structure in survey data. For example, we have applied this method to attitude data collected in the first wave of the COVID pandemic in the UK (Maher, P. J., MacCarron, P., & Quayle, M. (2020). Mapping public health responses with attitude networks..... British Journal of Social Psychology, 59(3), 641–652. https://doi.org/10.1111/bjso.12396) and American political attitudes (Dinkelberg, A., O’Reilly, C., MacCarron, P., Maher, P. J., & Quayle, M. (2021). Multidimensional polarization dynamics in US election data..... Analyses of Social Issues and Public Policy, 21(1), 284–311. https://doi.org/10.1111/asap.12278). After deriving a graph from the survey data, we can apply network methods to detect polarization (even without extremism); identify items which most clearly distinguish groups; and track polarization over time. How much overlap is there between groups? How can we identify polarization across multiple opinion dimensions? The ERC-DAFINET project provides a breakthrough mathematical method to answer these questions, but previously required high levels of mathematical and coding expertise to implement from first-principles.
The Attitude-Maps-4-All Proof of Concept project prototyped open software for implementing this network method, making it available to ordinary end-users (researchers and practitioners across disciplines). Where it previously required high-level technical mathematical and coding skills to implement, as a result of the work done in this Proof of Concept grant, it is now accessible to any researcher.
This Proof of Concept project developed a blueprint for translating the technical advance achieved in the ERC project into a form accessible and useable in the multidisciplinary scientific community, and laid the groundwork for the user and development community to make this a success.
This project had three dimensions. First, the IP and knowledge transfer workplan developed an optimal IP strategy for knowledge transfer to maximize the social impact of the advance.
Second, The technical dimension prototyped architecture and code immediately useable in commonly used statistical and data analysis software (C++; R; Python). These prototypes were tested, validated and revised with feedback from the early-adopter community (discussed below).
Third we prototyped documentation, tutorials, and training materials; and began to develop and train a community of early adopters. Early adopters were supported in using the prototype software to apply the method to their own interests and thereby validate it across a range of application areas.
This project had three dimensions. First, the IP and knowledge transfer workplan developed an optimal IP strategy for knowledge transfer to maximize the social impact of the advance.
Second, The technical dimension prototyped architecture and code immediately useable in commonly used statistical and data analysis software (C++; R; Python). These prototypes were tested, validated and revised with feedback from the early-adopter community (discussed below).
Third we prototyped documentation, tutorials, and training materials; and began to develop and train a community of early adopters. Early adopters were supported in using the prototype software to apply the method to their own interests and thereby validate it across a range of application areas.
The method for converting survey responses to homophily networks developed in the DAFINET ERC project is computationally intensive for large systems (since all dyads must be compared; and computation time therefore scales exponentially). Therefore, in the Attitude-Nets-4-All Proof of Concept project, the core of the software prototype was written in C++ for maximum efficiency and longevity, and is approximately 150 times faster than a similar algorithm scripted in R. It also should have a long shelf-life, since the C++ executable can be accessed directly. This C++ functionality is exposed to R and Python users via bespoke packages written for maximum efficiency and minimal dependencies, meaning that they should remain usable through many update cycles.
At the time of writing, the package is available to the global research community on GitHub via https://surveygraph.ie and in directly in R via Cran (package:surveygraph). The package is currently being used by early adopters. Training were run, and materials made available online. Self-guided workshop materials in the form of Jupyter Notebooks were made available.
At the time of writing, the package is available to the global research community on GitHub via https://surveygraph.ie and in directly in R via Cran (package:surveygraph). The package is currently being used by early adopters. Training were run, and materials made available online. Self-guided workshop materials in the form of Jupyter Notebooks were made available.