Network Footprints: An Experiment on Brokerage and Information Diffusion

How to cite

Renzini, Francesco; Squazzoni, Flaminio. 2023. "Network Footprints: An Experiment on Brokerage and Information Diffusion." AEA RCT Registry. AEARCTR-0011513. https://doi.org/10.1257/rct.11513-1.1

Download: BibTeX · RIS — for reference managers without a browser translator. Zotero and most others pick up this page's citation metadata directly.

Francesco Renzini* , Flaminio Squazzoni

* Primary investigator

Registry ID
AEARCTR-0011513
DOI
https://doi.org/10.1257/rct.11513-1.1
First registered
Last updated
Status
In development
Trial dates
2023-06-14 – 2023-11-15
Countries
Italy (Milan, Lombardy)
JEL codes
C9, C91
Registry record
http://www.socialscienceregistry.org/trials/11513

Abstract

Computational research on social networks has recently made progress in the understanding of how particular structural patterns emerge from individuals' tie-selection rules. It also has identified certain key network positions for the spread of epidemics and information. However, these models rely on idealized behavioral assumptions, such as perfect network information and awareness of the relevance of one's network position, which require systematic empirical test. Here, cognitive research on social network representation and memory has shown that individuals tend to have inaccurate perceptions and representations of the patterns of connectivity in their own social groups. Unfortunately, this research does not consider whether individuals--i.e., network brokers--who occupy particular network positions are aware of their own structural relevance for diffusion.

In this laboratory experiment, we aim to understand whether individuals who i) connect disconnected others in their personal networks and ii) have direct experience as network brokers are more likely to choose the network broker as the best starting point for a diffusion process in an artificial network. Subjects will be students and their previous experience as network brokers will be estimated before the experiment from ego network data collection and from a 16-item questionnaire, 12 of which were used in previous research. In the treatment group, subjects will be asked to choose through which node they expect information will transmit to most other nodes in the shortest time. In the control group, they will observe the same network, but had to select the node that is expected to spread an infection to the whole network in the shortest time. In both groups, subjects will be exposed to two compelling nodes: A "hub," which is a highly connected node that serves as a red herring; and a broker, which connects two dense communities. We expect that in the control group, there will be no difference between brokers and non-brokers choices. In the treatment group, we expect that brokers will correctly select the broker, while non-brokers will more frequently choose the red herring. Drawing from the network cognition literature, we expect that subjects with previous experience as network brokers will be aware of the structural importance of brokers in spreading information, but not in case of virus diffusion, which is a non-intentional process.

Experimental design

Subjects enter the lab in different sessions and sit at 24 separate workstations without the possibility to see each other's screen.

They open a web page, as the experiment was developed in html, css and javascript.

First they see an instruction page, which is also read aloud by the experimenter. They give their consent to proceed.

Then the phase of collecting ego networks begins, as they enter the names of family members, friends (made before and during university), work colleagues, roommates, members of associations or people with whom they share a hobby, and other people important to them, on successive web pages.

At the end of the ego network collection, we ensure that there is no error, and we present them with each potential pair among their contacts and ask them to check a box to indicate whether the two people would interact with each other even in the absence of the respondent.

They perform an attention check and task to make sure they are still paying attention before asking brokerage and personality questions.

They are asked to select on a series of sliders their response to the 16-item brokerage scale, a control question, and the personality questions on various Web pages.

They will be shown a network to familiarize them with the concept. We will make sure everyone knows what connections mean and imply by completing a 2-item visual questionnaire and another attention check.

Finally, they will select their seed from a dropdown menu in a network where a node and broker are clearly visible. The time remaining to select a node is displayed in red (45-second counter, with an alarm alerting them that 20 seconds remain).

After the treatment, some further data is collected. In particular, some sociodemographics (e.g., sex, whether they live in Milan or not, their enrollment year), whether their parents ever forced them to start an extracurricular activity, and whether they believe they were able to establish most of their relationships autonomously or with the help of someone else.

Design details

Randomization method
randomization done in office by a computer
Randomization unit
Unit (individual). We will randomize students in the two groups
Clusters
1
Observations
250 students from the university of Milan
Arms
approximately N/2 students in treatment and N/2 students in control groups
Primary outcomes
The dependent variable we are interested in is the seed selection of each participant. Thus, we want to measure which node a particular participant selected. This will be a binary variable. 1 for the network broker, 0 for the red-herring hub (or any node other than the broker; however, we expect very few nodes other than the broker and the hub to be selected). Therefore, we want to explain the probability of selecting the broker as a function of personal experience with network brokerage, in both the treatment and control groups.
Secondary outcomes
We will also measure subjects' personality, since it is an antecedent of brokerage behavior and may also affect seed selection. We will use a short version of the standard Big-Five personality test with acceptable psychometric properties for research that requires a quick and reliable measure (Gosling et al., 2003; A very brief measure of the Big-Five personality domains; Journal of Research in Personality). We will also ask more specific questions about "chameleonic" social behavior (see, e.g., Grosser et al., 2019) and other traits that may play a confounding role (see, e.g., Kalish and Robins, 2006, Psychological predispositions and network structure: The relationship between individual predispositions, structural holes and network closure, Social Networks).

Cited by

No citation data available for this registration yet.

Keywords