"Don’t Mess Up My Algorithm": Phatic Communication and Algorithmic Contagion in Meme Sharing
"Don't Mess Up My Algorithm": Phatic Communication and Algorithmic Contagion in Meme Sharing
HCI Today summarized the key points
- •This article discusses research on how meme sharing in Instagram DMs affects relationship maintenance and users’ perceptions of recommendation algorithms.
- •The research team interviewed 21 active meme DM users in South Korea and analyzed meme sharing as a form of social communication for confirming intimacy, rather than as information transfer.
- •Participants categorized memes as friendly or unfriendly depending on their fit with the recipient, and they perceived the spread of unfriendly memes into the feed as algorithmic contamination.
- •However, feedback such as ‘Not Interested’ was less effective, and blocking or ending the conversation was seen as damaging the relationship—so many users effectively could not respond.
- •The study suggests that regaining users’ sense of control requires separating DM-based relationship maintenance from recommendation learning, explaining the DM linkage, and using more conservative learning.
This summary was generated by an AI editor based on HCI expert perspectives.
Why Read This from an HCI Perspective
This article is highly relevant for HCI/UX practitioners because it shows how a private interaction like DM becomes entangled with recommendation systems—and how users interpret and try to avoid that entanglement. In particular, it reveals the burden and sense of powerlessness users feel when ‘maintaining relationships’ conflicts with ‘algorithmic control.’ It also explains why design must consider relational context, not just feed control.
CIT's Commentary
From a CIT perspective, the interesting point is that this study reframes ‘personalization’ not as a problem of information accuracy, but as a problem of relational cost. Users may interpret even the light act of exchanging notes as an algorithmic signal, and as a result, interactions that help preference learning can be experienced as feed contamination (contagion). This suggests that simply increasing the visibility of feedback buttons in UX is not enough; we also need to design, together, conversation-level separation, opt-outs that do not harm relationships, and the privacy precision of explanations. However, since this paper focuses on perceived mechanisms (folk theory), follow-up measurement studies that verify differences from the actual recommendation logic are needed to set the design recommendations’ priorities more accurately.
Questions to Consider While Reading
- Q.How understandable to users is a conversation-level opt-out that prevents DM-based interactions from being reflected in recommendations, and how much can it reduce relational awkwardness?
- Q.How can we distinguish whether content types perceived as ‘algorithmic contagion’ are a matter of individual taste, or whether they vary depending on social context and relationship strength?
- Q.What information design principles could implement a balance where an explanation interface transparently shows the DM–recommendation linkage while not exposing the sender or relationship information?
This commentary was generated by an AI editor based on HCI expert perspectives.
Please refer to the original for accurate details.
Subscribe to Newsletter
Get the weekly HCI highlights delivered to your inbox every Friday.