Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

The Atlas of AI Incidents in Mobile Computing: Visualizing the Risks and Benefits of AI Gone Mobile

Edyta Bogucka, Marios Constantinides, Julia De Miguel Velazquez, Sanja Šćepanović, Daniele Quercia, Andrés Gvirtz

TL;DR

The paper tackles the public-understanding gap in AI risks and benefits for mobile computing by building a narrative, interactive visualization around 54 real-world incidents drawn from the AI Incident Database. It advances a reproducible workflow (data-format design, incident curation, GPT-4.0-based paraphrasing, expert validation, EU AI Act and SDG assessments) and a Martini Glass visualization implemented with SBERT/t-SNE embeddings and D3.js to present risk levels and potential benefits. Key contributions include a structured, non-expert-friendly presentation of complex socio-technical risk, a framework linking regulatory (EU AI Act) and sustainable development (SDGs) perspectives, and a public tool for education and policy discussion. The work demonstrates that even ostensibly low-risk mobile AI uses can generate meaningful incidents, underscoring the need for accessible risk communication and informed design in mobile AI deployment.

Abstract

Today's visualization tools for conveying the risks and benefits of AI technologies are largely tailored for those with technical expertise. To bridge this gap, we have developed a visualization that employs narrative patterns and interactive elements, enabling the broader public to gradually grasp the diverse risks and benefits associated with AI. Using a dataset of 54 real-world incidents involving AI in mobile computing, we examined design choices that enhance public understanding and provoke reflection on how certain AI applications - even those deemed low-risk by law - can still lead to significant incidents. Visualization: https://social-dynamics.net/mobile-ai-risks

The Atlas of AI Incidents in Mobile Computing: Visualizing the Risks and Benefits of AI Gone Mobile

TL;DR

The paper tackles the public-understanding gap in AI risks and benefits for mobile computing by building a narrative, interactive visualization around 54 real-world incidents drawn from the AI Incident Database. It advances a reproducible workflow (data-format design, incident curation, GPT-4.0-based paraphrasing, expert validation, EU AI Act and SDG assessments) and a Martini Glass visualization implemented with SBERT/t-SNE embeddings and D3.js to present risk levels and potential benefits. Key contributions include a structured, non-expert-friendly presentation of complex socio-technical risk, a framework linking regulatory (EU AI Act) and sustainable development (SDGs) perspectives, and a public tool for education and policy discussion. The work demonstrates that even ostensibly low-risk mobile AI uses can generate meaningful incidents, underscoring the need for accessible risk communication and informed design in mobile AI deployment.

Abstract

Today's visualization tools for conveying the risks and benefits of AI technologies are largely tailored for those with technical expertise. To bridge this gap, we have developed a visualization that employs narrative patterns and interactive elements, enabling the broader public to gradually grasp the diverse risks and benefits associated with AI. Using a dataset of 54 real-world incidents involving AI in mobile computing, we examined design choices that enhance public understanding and provoke reflection on how certain AI applications - even those deemed low-risk by law - can still lead to significant incidents. Visualization: https://social-dynamics.net/mobile-ai-risks
Paper Structure (6 sections, 3 figures)

This paper contains 6 sections, 3 figures.

Table of Contents

  1. Introduction
  2. Related Work
  3. Visualizing the Risks and Benefits of AI Uses
  4. Curating a Dataset of Risks and Benefits
  5. Visualizing the Risks and Benefits
  6. Discussion and Conclusion

Figures (3)

  • Figure 1: Our six-step process for visualizing AI risks and benefits. We started with designing the data format required to visualize AI uses (Step 1). We then selected mobile computing uses associated with incidents in the AI Incident Database mcgregor2021preventing (Step 2) and placed them in our data format (Step 3). Next, we reviewed the correctness of the use formatting (Step 4), manually assessed the risks and benefits of the resulting uses (Step 5), and finally, visualized these uses (Step 6).
  • Figure 2: The visualization of AI risks introduces information through the narrative structure of a Martini Glass (shown at the top) unfolding in four sections (shown at the bottom). The first three sections, author-driven, introduce the dataset of uses, risks, and benefits. The final section, user-driven, supports dataset exploration tasks through an interactive dashboard.
  • Figure 3: The dashboard supports four dataset exploration tasks. Users can browse the use map with tooltips (T1), view detailed descriptions of uses by clicking on dots (T2), filter them by categories (T3), and find similar uses through keyword search (T4).