the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 20 Nov 2023
The Pareto effect in tipping social networks: from minority to majority
Abstract. How do social networks tip? A popular theory is that a small minority can affect network, or population wide change. This effect is roughly consistent with the properties of the Pareto principle, a semi-quantitative law which suggests that in many systems, 80 % of effects are produced by only 20 % of the causes. In the context of the transition to net-zero emissions, this vital 20 % can be a critical instigator of social tipping, a process which can rapidly accelerate social norm change. In this work, we ask whether the Pareto effect can be observed in social systems by conducting a literature review with a focus on social norm diffusion and complex contagion on social networks. By collecting simulation and empirical results of social tipping events over a wide disciplinary, and parametric space, we are able to see the existence of shared behaviour across studies. Based on a compiled dataset, we show general support for the existence of a tipping point which occurs at around 25 % of the total population in susceptible social systems. Around this critical mass, there is a high likelihood of a social tipping event, where a large minority is then quickly “tipped”. Additionally, we were able to show a range of critical masses where social tipping is possible, these values lie roughly between 10 % and 45 %. Finally, we also provide practical advice for facilitating norm changes under uncertainty, difficult social norm transitions, and social groups resistant to change.
- Preprint
(1627 KB) - Metadata XML
- BibTeX
- EndNote
Status: open (until 08 May 2024)
-
CC1: 'Comment on egusphere-2023-2241', Sibel Eker, 08 Feb 2024
reply
The Pareto effect in tipping social networks: from minority to majority
This manuscript presents a literature review on social norm diffusion/contagion on social networks, and whether a critical mass similar to a Pareto effect is observed. The manuscript makes a relevant and important contribution to the expanding literature on social tipping by aggregating the available empirical evidence on tipping dynamics. I also appreciate that authors clarify and describe tipping-related concepts and terms in Section 2, yet I believe that manuscript would benefit from a more condensed version of it.
My main concern is how “social tipping” is implicitly defined in the manuscript, and that it is not clearly aligned with the real-life examples of what social tipping in the climate change and sustainability context entails. For instance, the manuscript focuses on social networks and complex contagion, but does it mean that all social tipping process, such as rapid reduction of EV battery costs, coverage of climate change in the school curricula, as identified in the authors’ own paper from 2020, have to include contagion dynamics through a social network? Even if we can rightfully represent many of those in networks, do we have to? Is the presence of a network structure necessary condition for tipping? How does this relate to the existing social tipping and “positive” tipping literature that does not necessarily focus on networks? Answering such questions would improve the manuscript.
Relatively minor comments:
Lines 75-88 : This paragraph does a fair job in listing the objectives, yet they sound a bit too ambitious. For instance, “we identify critical factors influencing tipping processes in social systems” is eventually only limited to a handful of studies included in your review on social networks, therefore I suggest to reformulate these objectives more consistently with the actual methods and results.
Figure 1a: I don't find this figure very clear especially considering the accompanying text. If the “shaded area represents the domain where the system exhibits a nonlinear response” would any nonlinear curve above the y=x line mean "tipping"? How about the zone below the line? It might also include very nonlinear changes.
Line 121-122: “where roughly a 20% change in a system control parameter results in an 80% change in the system state at equilibrium”. I am not sure if this is the right formulation of your idea. What happens as lambda moves from 0.1 to 0.12, which is a 20% change in the control parameter?
Line 126-127: “the shown Pareto CDF depicts a scenario in which a minority of actors have convinced a large majority to switch to another social norm.” I think it doesn't show that one, because the y axis is F when t goes to infinity, the final state. When lambda=0.8, we end up at >0.8 and no tipping occurs. The pareto cdf line covers many different scenarios not necessarily those where a minority convinces the others.
Figure 1b: Appreciating this plot of tipping dynamics over time, yet it is the typical s-curve of transitions/innovation adoption. Could you expand on what makes it “tipping”? Furthermore, it is not clear on the figure what the tipping threshold is, as suggested in Box 1.
Line 182: Using cascade and social tipping interchangeably. “Cascade” has been used (in the GTPR) recently to refer to the tipping dynamics in different systems affecting each other, both for climate tipping points and positive tipping points. This terminology choice might be confusing for the readers.
Figure 3: Not clear what the x- and y- axes represent.
Line 368: 21 papers were discarded because they did not include complex contagion. It would be useful to mention what they included instead.
Table 4: Very useful table summarizing the reviewed studies. To strengthen the conclusions of the manuscript, we need to know more about these studies, though. Could you specify sample size/network size, geographic location, what the nodes and edges are, socioeconomic and demographic background etc.?
Figure 6: Very informative figure especially because it distinguishes between model-based and empirical studies! The sample size is confusing, though, since the caption says that it includes 87 papers and the text earlier mentioned the eventual # of studies in the review were ~30. Please clarify. It raises the question of what made the empirical studies with a similar lambda around 20% but not tipped (bottom left). This figure could have been improved in a bubble chat format, where the dot size refers to the network size in each study. “Population size” came up as one of the most important drivers in Fig.4, and we do observe its relevance. For instance, the global percentage of vegetarian population is 20%, but no tipping is observed.
Line 518- Evidence of critical mass: For the empirical studies covered in this paper, it would be very informative for the strength of evidence whether they are in lab settings or in a real-life experimental setting or contain large-scale data?
Line 531-533 “For modellers…” : Could you expand on “validation across modelling approaches” for computational efficiency? There are examples of system dynamics models which do not use a complex contagion and threshold approach, but show that the inflection point of the norm adoption function (the counterpart of the cdf of the probability of an agent adopting a norm for the fraction of agents who already adopted it) is the most important driver of large scale change in the diet context. How do we for instance cross-validate those? https://www.nature.com/articles/s41893-019-0331-1
Line 542: Undermining what you did.
Minor comments:
Line 45 “mechanisms. In which…” please watch the grammar.
Line 174 Please clarify what “systems-dynamics” is, since it is not a commonly used term.
Line 338: Which “Ref”?
I suggest to put table captions above the table, since it is a more commonly used convention.
Line 497: table x
The first paragraph of Section 5: I suggest to divide this paragraph and several others, since they are too lengthy and contain more than 1 argument, main point etc.
Citation: https://doi.org/10.5194/egusphere-2023-2241-CC1 -
RC1: 'Comment on egusphere-2023-2241', Anonymous Referee #1, 15 Apr 2024
reply
This is a very interesting work that combines a challenging literature review with some results. I say "challenging" because the topic (tipping in social networks) demands several expertises that can be quite far from each other. Network science and social network analysis, dynamical systems (phase transitions, non-linear dynamics, etc.), socio-ecological systems, simulation, data analysis, and so on.
The manuscript makes a good job at introducing the core concepts needed for such discussion, and the effort to integrate all those, coming from different disciplines, is generally successful: someone unfamiliar with social networks, information cascades, etc. will achieve a sufficient understanding of the research problem.
Beyond this general positive impression, the manuscript suffers from some shortcomings that should be addressed. In some sense, these weaknesses are "technical" (writing style, typos), but they end up affecting the message, and not only the form. For example, we can find some sentences which are unfinished or out of context (example: line 210: "Although in the former it is the spread of cooperation.". It makes reference to the previous text, but this sentence ill-formed: it reads as if something had to follow after 'cooperation'. Yet another example (line 260): "An example is reducing" (the sentence is unfinished, it simply ends there).
I am aware that these are easy-to-solve issues. There are more (perhaps not so obvious). My point is that such formal errors interfere with the smoothness of the text --which needs to be really smooth if meant as a review for scholars from other disciplines. These errors (and many other smaller ones) suggest that the text was written with some hastiness.
A second indication of hastiness is the fact that some paragraphs are redundant (they can be some pages apart). I think that the concept of "threshold" is a good illustration of this: both their macro and individual dimensions keep appearing in different places, and so such a central concept is at risk of becoming "fuzzy" (except for the explicit and clear definitions in Box 1). My suggestion is then to reconsider the style and organization, specially in Sec. 2, to optimize the pedagogical value of the work.
I recommend caution as well when using certain words. In Box 1, we see "percolation" and "spreading" as equivalent (verbatim: "[...] which after a percolation (spreading) process must occupy [...]". Although percolation is behind the study of contagion-like processes on networks (disease, information), it has a wider meaning in network science and statistical physics.
For the sake of transparency and reproducibility, details about data collection (Sec. 3.1) should be more precise. The literature search must have for sure retirned hundreds, if not thousands, of titles. How then the authors reached an initial number of 33?
To sum up, the value of a document like this lies in its clarity. Precisely because the scientific content of the manuscript is very appealing, the effort to communicate it must be equated.
Finally, other minor issues to take into account:
- The "Clustered lattice" representation in Fig. 2 is very counterintuitive. I am sure a better representation is possible.
- Table 1: Erdös-Rényi networks have a low average path length.
- Figure 3: please add labels to x- and y-axis
- Please revise the references format. One can find, e.g., "PNAS", "Proceedings of the National Academy of Sciences"; missing publication years; etc.Citation: https://doi.org/10.5194/egusphere-2023-2241-RC1
Viewed
| HTML | XML | Total | BibTeX | EndNote | |
|---|---|---|---|---|---|
| 263 | 101 | 16 | 380 | 12 | 11 |
- HTML: 263
- PDF: 101
- XML: 16
- Total: 380
- BibTeX: 12
- EndNote: 11
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1