Habitual Sharing

What it is

The paper starts from a question about explanations. Sharing of false news is often put down to lazy reasoning, to bias, or to both. Ceylan, Anderson and Wood asked whether a third factor does more of the work: the habit of sharing itself. In habit research a habit is a response that a context triggers automatically after the response has been rewarded many times. The authors apply that idea to the act of passing a headline on.

Their account rests on reward learning. People learn to share content that attracts attention, because attention is the reward the setting returns. Once the habit has formed, cues in the setting trigger sharing with little regard for whether the content is true. On this account the strength of the habit, and not the quality of a person's reasoning at the moment of choice, predicts a good deal of what gets shared.

To test it, the authors had participants decide whether to share 16 simulated headlines, some true and some false. They measured habit strength in two ways: self-reported automaticity of sharing, and how often the person had shared in the past. They could then relate habit strength to how many headlines a person shared and to discernment, meaning how well the person's sharing separated true headlines from false ones. The idea sits close to habit automaticity and to the way self-reported habit strength is measured in the habit literature.

In effect

Stronger habits went with sharing more headlines overall and with less discrimination between true and false headlines. The 15 per cent of participants with the strongest habits accounted for roughly 30 to 40 per cent of the false headlines shared, the exact figure depending on how habit was measured. Habitual sharers also stayed insensitive to accuracy when they were asked to judge it before deciding, so a prompt to think about accuracy did little for the strongest habits.

The fourth study moved from description to training. Participants who were rewarded for sharing accurate news became better at sharing true over false headlines, and the improvement carried over to later sessions in which no rewards were given. Participants trained with rewards for sharing misinformation lost discernment. Accuracy training did not reduce how often people shared, and it slightly raised it.

Read alongside accuracy inattention, the result gives a second reason why accuracy does not govern sharing. In the accuracy studies, attention was elsewhere at the moment of choice. In this study the response had become automatic, so the choice was barely made at the moment at all. The authors conclude that the sharing structure of social media shapes behaviour more than individual reasoning or bias does, and that it could be designed to build habits of sharing accurate information.

What it does not say

The study does not say that all sharing is habitual, or that reasoning and bias play no part. It reports that habit strength explained a good deal, and that a small group of strong habits accounted for a large share of false headlines in its samples.

The headlines were simulated, and habit strength came from self-report and from reported past sharing. The figures describe these samples and these measures. The training result is an experimental finding about sessions run by the researchers, and it is not a forecast of what would happen if rewards for accuracy were built into a real setting.

The 30 to 40 per cent figure is also easy to misread. It describes how false headlines shared were distributed across participants in the study. It does not describe how much false content exists, how much people see, or how many people share it in daily life.

Who developed it

Gizem Ceylan, then listed at Yale University, is the first author. Ian Anderson and Wendy Wood are listed at the University of Southern California. Wood is a habit researcher whose wider work is covered on her own page, Wendy Wood. The concept is named here from the paper's central claim, that sharing of misinformation is habitual, and the authors' own account is the one set out above.

---