Analysis of AI-Driven Surveillance in Domestic Work

"These cameras are just like the Eye of Sauron"

This research explores the intersection of artificial intelligence, smart home technology, and the precarious nature of domestic labor. The study focuses specifically on how UK-based domestic workers (DWs) perceive and experience the privacy threats posed by AI-integrated devices.

📄 Document Metadata

Authors:

• Shijing He
• Yaxiong Lei
• Xiao Zhan
• Ruba Abu-Salma
• Jose Such

---

🔍 Research Overview

The proliferation of AI-powered smart home devices has created a complex privacy landscape for domestic workers. These individuals occupy a unique position: they are monitored by AI in the homes of their employers and manage AI devices within their own private residences.

Methodology

The researchers employed a human-centered approach to threat modeling, utilizing:

1. Semi-structured interviews with a cohort of 18 domestic workers in the UK.
2. Communication Privacy Management (CPM) as the theoretical framework to analyze how privacy boundaries are negotiated.

The Sociotechnical Flow

The study moves beyond simple hacker-based threats to look at systemic risks. The following diagram illustrates the data and surveillance flow identified in the paper:

---

🚩 Key Findings

1. In the Employer's Household

In these environments, the worker has the least agency. The research highlights:

• Intensified Surveillance: AI-enabled features (e.g., motion alerts, facial recognition) act as a constant "Eye of Sauron."
• Institutional Barriers: Employment agencies often mediate the relationship, making it nearly impossible for workers to negotiate privacy boundaries.
• Data Persistence: The risk is not just real-time viewing, but residual data logs and cross-household data flows.

2. In the Worker's Own Household

While the worker is the owner here, privacy is still compromised by:

• Administrative Burden: A "gendered" division of labor where certain individuals manage the tech but don't fully understand its implications.
• Technical Opacity: Uncertainty regarding how AI functions and how long data is retained.

3. The Threat Model

The authors propose a new sociotechnical threat model. If we were to represent the perceived risk mathematically, it might look like this:

$\text{Privacy Risk} = \frac{(\text{AI Analytics} + \text{Data Retention}) \times \text{Institutional Power}}{\text{User Agency}}$

---

🛠️ Implications and Goals

The study concludes by identifying domestic worker agencies as institutional adversaries—entities that may prioritize monitoring over worker privacy.

Future Objectives for Privacy Strengthening:

• Increase transparency of AI functionalities in consumer devices.
• Establish clear legal boundaries for monitoring domestic staff.
• Develop tools for workers to audit the data being collected about them.
• Address the gendered gap in technical administration within the home.

---

Access Links:

• View PDF | HTML (experimental) | TeX Source
• DOI: https://doi.org/10.48550/arXiv.2602.09239