Impressively scary - Exploring user perceptions and reactions to unraveling machine learning models in social media applications

Aug 3, 2026

Upload by Aurelia

Machine learning models deployed locally in social media applications enable functionality such as real-time face filters, yet expose sensitive attributes while remaining difficult for users to understand. The study investigates how transparency about when and how such models are used affects user perceptions and behaviors. A user study (N=21) shows participants are often unaware of model outputs and usage timing in Instagram and TikTok. Exposure leads to long-term behavior changes in 8 participants, highlighting challenges and opportunities for ML transparency using local user data.

LanguageEnglish

CategoryResearch & Report

FormatPDF

Pages22

Reading Time55 min

AccessFree to download

Introduction

Motivation and market context

Opacity of AI/ML deployment

Study and findings

User study design (N=21)

Awareness of outputs and usage timing

Long-term behavior changes

Challenges and opportunities

Transparency for local user-data ML

Edinburgh Research Explorer “Impressively scary" Exploring user perceptions and reactions to unraveling machine learning models in social media applications Citation for published version: West, J, Cagiltay, B, Zhang, S, Li, J, Fawaz, K & Banerjee, S 2025,“Impressively scary": Exploring user perceptions and reactions to unraveling machine learning models in social media applications. in N Yamashita, V Evers, K Yatani, X Ding, B Lee, M Chetty & P Toups-Dugas (eds), CHI '25: Proceedings of the 2025 CHI Conference on Human Factors in Computing Systems. , 818, Association for Computing Machinery (ACM), New York, NY, United States, pp. 1-21, 2025 Conference on Human Factors in Computing Systems, Yokohama, Japan, 26/04/25 . [https://doi.org/10.1145/3706598.3713256](https://doi.org/10.1145/3706598.3713256) Digital Object Identifier (DOI): 10.1145/3706598.3713256 Link: Link to publication record in Edinburgh Research Explorer Document Version: Publisher's PDF, also known as Version of record Published In: CHI '25 General rights Copyright for the publications made accessible via the Edinburgh Research Explorer is retained by the author(s) and / or other copyright owners and it is a condition of accessing these publications that users recognise and abide by the legal requirements associated with these rights. Take down policy The University of Edinburgh has made every reasonable effort to ensure that Edinburgh Research Explorer content complies with UK legislation. If you believe that the public display of this file breaches copyright please [contact openaccess@ed.ac.uk](contact openaccess@ed.ac.uk) providing details, and we will remove access to the work immediately and investigate your claim. Download date: 08. Jan. 2026 “Impressively Scary:’ Exploring User Perceptions and Reactions to Unraveling Machine Learning Models in Social Media Applications Jack West Department of Computer Sciences University of Wisconsin-Madison Madison, Wisconsin, USA [jwwest@wisc.edu](jwwest@wisc.edu) Bengisu Cagiltay Department of Computer Sciences University of Wisconsin-Madison Madison, Wisconsin, USA [bengisu@cs.wisc.edu](bengisu@cs.wisc.edu) Shirley Zhang Department of Computer Sciences University of Wisconsin-Madison Madison, Wisconsin, USA [hzhang664@wisc.edu](hzhang664@wisc.edu) Jingjie Li School of Informatics University of Edinburgh Edinburgh, United Kingdom [jingjie.li@ed.ac.uk](jingjie.li@ed.ac.uk) Kassem Fawaz Electrical and Computer Engineering University of Wisconsin-Madison Madison, Wisconsin, USA [kfawaz@wisc.edu](kfawaz@wisc.edu) Suman Banerjee Department of Computer Sciences University of Wisconsin-Madison Madison, Wisconsin, USA [suman@cs.wisc.edu](suman@cs.wisc.edu) Abstract Machine learning models deployed locally on social media applications are used for features, such as face filters which read faces in-real time, and they expose sensitive attributes to the apps. However, the deployment of machine learning models, e.g., when, where, and how they are used, in social media applications is opaque to users. We aim to address this inconsistency and investigate how social media user perceptions and behaviors change once exposed to these models. We conducted user studies (N=21) and found that participants were unaware to both what the models output and when the models were used in Instagram and TikTok, two major social media platforms. In response to being exposed to the models’functionality, we observed long term behavior changes in 8 participants. Our analysis uncovers the challenges and opportunities in providing transparency for machine learning models that interact with local user data. CCS Concepts • Security and privacy → Social aspects of security and privacy; • Human-centered computing → Empirical studies in HCI. Keywords AI Transparency; Social Media; Privacy; Mobile AI ACM Reference Format: Jack West, Bengisu Cagiltay, Shirley Zhang, Jingjie Li, Kassem

FAQs

What problem does the paper address about machine learning models in social media?

It addresses the mismatch between ML functionality that uses sensitive local data and users’ limited understanding of when, where, and how those models are used.

How did the authors evaluate user perceptions and reactions?

They conducted user studies with 21 participants to examine awareness of what models output and when models were used in Instagram and TikTok.

What were the main behavior outcomes after users were exposed to model functionality?

The study observed long-term behavior changes in 8 participants after exposure, demonstrating measurable effects of increasing transparency.

Aug 3, 2026

Upload by Aurelia