(Last Updated: April 29, 2021)
One of Tattle's goals is to archive viral content from closed messaging apps such as WhatsApp, Signal and Telegram. In entering groups on closed messaging apps, researchers and developers must be mindful to not treat these groups as public by default. People sign up on messaging apps using their phone numbers which could identify them. Surfacing data from these groups, without adequate concerns about user safety, could put chat app users at risk.
This is an evolving check-list of design decisions that must consider before beginning any data collection from chat apps. The design choices in setting up the data collection process from closed messaging apps can be broken down into five main stages.
For a more detailed background on this check-list and description of these steps, please see the paper:
Sehat, C. M., Prabhakar, T., Kaminski, A. (2021, March 15). Ethical Approaches to Closed Messaging Research: Considerations in Democratic Contexts. Retrieved from https://electionstandards.cartercenter.org/verifying-elections-misinfocon2020/ethical-approaches-to-closed-messaging-research-considerations-in-democratic-contexts/
The phone number could be procured by researchers as individuals or by an organization. Phone numbers can be seen by everyone in a group conversation and might be used for digital marketing and other unsolicited communication. Separating phone numbers used for research purposes and for personal communication can add the necessary separation between researchers personal lives and the research work.
At the time of signing up on a messaging platform, a user has to declare a name that identifies them to others on the app. Researcher(s) may choose to use individual names or the name of the research group/study. Declaring the name as a research study or organization might make it easier for group admins and participants to identify the researcher, enabling more informed and meaningful consent in the admission of researchers in the group.
Researchers might provide additional description of their purpose in the ‘Status’ or ‘About’ field, if the platform allows it. Using language that research subjects understand can result in more meaningful consent. This however may be difficult in multilingual groups. Limit on the length of the text description restricts lengthy or multi-lingual description.
Regardless of the username declared on the messaging app, note that the owner of a phone number may still be identified through apps such as True Caller. For researchers working in sensitive contexts, identification might be another consideration to have phone numbers procured through an organization.
On apps such as Telegram, “public channels” can be found through searching for the channel names on the app. WhatsApp, however, does not provide this option. Links to group conversations must be found by proactive searching for group links online. This search could span Facebook groups, Twitter conversations or a broad search on the world wide web. While Telegram aggregates popular channels on the app, several websites (not endorsed by WhatsApp) aggregate WhatsApp group links. Links to new group chats may also be shared within a group that researchers are monitoring.
The nature of the online forum on which the message group link is found, can provide some indication if the groups were inteded to be public. Message groups shared exclusively on closed Facebook groups, for example, should be assumed to be used by members of a closed community. Group links shared on public Twitter profiles could be created for broader engagement.
With group links aggregated by independent websites, the intended nature (public vis-a-vis private) of the group is hard to discern. The selection and categorization of groups on a website is decided by website creators and not the message group admins or participants. Thus, researchers must use their discretion in using these websites for research purposes.
Finally, researchers may consider joining additional groups through links shared on groups that they are monitoring. The public status of these groups is derived from the perceived status ascribed to the message group they were found on. If the assumption is that the public status of all messaging groups is unclear to begin with, the public status of groups found on other messaging groups is doubly weakened.
Disclosure Norms: On joining a conversation, researchers may choose to disclose their purpose. Doing so allows for more informed by admins and may result in researchers being removed from certain groups.
Deciding to Leave a Group: After joining a group, researchers may realize that the group is intended to be private. Group size maybe used as a proxy for private or public nature of the group. The number of people in a message group can only be found after joining it. Since the size of a group dynamically changes, researchers may choose to keep a time window after joining, before deciding to study the group or leave it. Thoughtful protocols may be followed in this time window to respect privacy and consent of group participants. Group settings such as 'disappearing messages', might imply that the group participants do not want their group conversations recorded.
Public or not, message groups are a space for consequential discourse. People may express political views or share their health histories. All the message conversations are tied to a phone number that could be used to link these conversations to other databases. Truly anonymous data collection may not be possible. It is imperative that at the very least the data is de-identified. Deidentification does not guarantee anonymization but as more data fields are de-identified, the risk of deanonymization decreases. In addition to phone numbers, researchers may also consider anonymizing the group conversation name. If the focus is only the message content, researchers may choose to use a different anonymization ‘seed’ for new batch of data. This would delink conversations from the same sender across different data snapshots.
The method adopted for extracting data, anonymizing it as well as analyzing data should be clarified in publications using the data. Researchers may also consider opening the source code used to extract and analyze the data. Documenting ethical practices of data collection in the documentation of the code could remind others who use the code to take adequate measures to protect the subjects they are studying.
In studying groups on closed messaging apps, researchers define some conditions under which the closed message groups are considered public enough to be studied. As the definition of public is expanded to include more groups or increase the granularity of data collection, the burden to ensure security of the data also increases. Researchers must gauge their ability to securely analyze and share the volume of data they collect in their research.
By codifying responsible data management practices prior to the study, researchers can anticipate and mitigate risks to their research subjects. In absence of an Institutional Review Board, researchers could have their research protocol reviewed by professional peers, not tied to the research project. Researchers could similarly also have their data collection and storage practices vetted by cybersecurity experts.
{/ Add projects here to show the blogs and updates /}