Corrigendum to 'Arguments as Drivers of Issue Polarisation in Debates Among Artificial Agents', Journal of Artificial Societies and Social Simulation, 25 (1) 4, 2022
DebateLab, Karlsruhe Institute of Technology, Germany
Journal of Artificial
Societies and Social Simulation 26 (2) 3
<https://www.jasss.org/26/2/3.html>
DOI: 10.18564/jasss.5068
Received: 01-Mar-2023 Accepted: 07-Mar-2023 Published: 31-Mar-2023
Abstract
This corrigendum refers to 'Arguments as Drivers of Issue Polarisation in Debates Among Artificial Agents', Journal of Artificial Societies and Social Simulation, 25 (1) 4, 2022.Correction 1
Figure 1 replaces the original Figure 3 in Section 4.7 in Kopecky (2022) as the original version reported results from incorrect measurement. In the original manuscript, the dispersion measurement was normalised incorrectly to a [0, 0.5] range. This was due to a factor 2 missing from Definition 1 in Section 4.6 of the original manuscript. Adding this factor to the definition yields measurements normalised to a [0, 1] range. The correct calculation for dispersion is as follows:
\[\text{dispersion}(\tau) := 2\Bigg(\frac{1}{N}\sum_{i\in A_\tau}^{N}\Bigg(\frac{1}{N-1}\sum_{j\neq i}^{N-1} \delta(i,j) - \underbrace{\frac{1}{\left|{A_\tau\choose 2}\right|} \sum_{(x,y)\in{A_\tau}} \delta(x,y)}_{\text{mean distance (constant)}} \Bigg)^2\,\Bigg)^{1/2}\] |
This implies that all dispersion measurements in the original manuscript must be doubled. The new figure shows that dispersion values are closer to the observed group divergence values than reported in the original manuscript. Comments made in the original article about dispersion values being at a lower level were biased by the erroneous normalisation. Indeed, dispersion appears to be a useful approximation of group divergence as it does not require community structuring.
However, as the findings of the original manuscript relied mainly on group divergence and group consensus measurements, the main conclusions of the original article are unaffected by this error, and even strengthened by the correct normalisation of dispersion values.
Correction 2
Figure 2 replaces the original Figure 13 included in the original Appendix C. Figure 2 contains raw dispersion values calculated with the correct normalisation.
Correction 3
The model code in the original article at this link: https://doi.org/10.5281/zenodo.5067834 has been updated with a new pairwise_dispersion()
function. The normalisation has been corrected from taupy
version 0.5 onwards.
The author would like to apologise for any inconvenience caused.
References
KOPECKY, F. (2022). Arguments as Drivers of Issue Polarisation in Debates Among Artificial Agents, Journal of Artificial Societies and Social Simulation. 25 (1) 4: http://jasss.soc.surrey.ac.uk/25/1/4.html. [doi:10.18564/jasss.4767]