Entry-point challenge: Actors' loyalty and temporal commitment in discourse networks

[leifeld]

This issue serves as an entry-point research challenge focused on the measurement and estimation of actors' loyalty (or temporal commitment) to concepts in discourse networks.

Challenge

Several existing approaches define actors' loyalty as a function of how consistently and recently actors support the same concepts (e.g., policy positions) over time. A prominent example is the loyalty measure proposed by Sharara et al. (2011), which combines frequency and recency using a user-specified temporal discounting parameter.

The challenge is to analyse, diagnose, and potentially extend such loyalty measures, with particular attention to:

• the role and interpretation of temporal decay parameters;
• identifiability and sensitivity of loyalty measures to tuning choices;
• whether decay or commitment parameters can be estimated from data rather than fixed a priori;
• how alternative formulations affect inferential interpretation in temporal bipartite networks of actors and concepts.

This challenge treats loyalty not as a descriptive score, but as a measurement and inferential design problem.

A full implementation is not required.

Possible outputs

Useful contributions may include, but are not limited to:

• a technical analysis of existing loyalty measures and their assumptions;
• simulation-based diagnostics exploring sensitivity to decay or window parameters;
• alternative formulations or estimators (e.g. Bayesian, optimisation-based);
• prototype implementations in R or Java;
• a short technical note clarifying interpretational or identifiability issues.

Partial and exploratory work is entirely appropriate.

Work on this challenge can be shared via a fork or pull request; in some cases, exploratory work may also live in a separate repository linked from this issue. Discussion and partial results can be documented directly here.

Background, prior work, and relation to DNA

Existing loyalty measures were introduced in Sharara et al. (2011; 2012). These approaches rely on fixed temporal discounting parameters, raising questions about robustness, interpretation, and estimation.

An earlier application is documented in Tolstukha (2022), which applies related loyalty ideas in a discourse network setting but does not address estimation or inferential identification directly. It can serve as a useful reference point for design decisions and pitfalls.

The challenge is motivated by ongoing work on measurement, temporal dependence, and estimation in discourse networks. Any resulting methods could, in principle, be integrated into DNA or rDNA, but integration is not expected at this stage.

References

• Sharara, H., Singh, L., Getoor, L., & Mann, J. (2011). Understanding actor loyalty to event-based groups in affiliation networks. Social Network Analysis and Mining, 1(2), 115-126. [link]
• Sharara, H., Singh, L., Getoor, L., & Mann, J. (2012). Stability vs. Diversity: Understanding the Dynamics of Actors in Time-Varying Affiliation Networks. 2012 International Conference on Social Informatics. 14--16 December 2012. [link]
• Tolstukha, E. (2022). Stalemate in the democratic reform debate of the European Union? A dynamic discourse network analysis of actors and their commitment to reform options. PhD thesis, School of Social and Political Sciences, University of Glasgow. [link]