Generative Ghosts: Anticipating Benefits and Risks of AI Afterlives

Generative AI (Text, Image, Music, Video)Algorithmic Fairness & BiasTechnology Ethics & Critical HCIAI/ML Researchers & EngineersHCI ResearchersSociologists & Anthropologists

Research Background and Problem

What issues or challenges did the authors identify?

The authors explore the rapid advancements in generative AI systems, which may lead to the common practice of interacting with the deceased through customized AI agents in the future. These systems, referred to as "generative ghosts," are capable of creating new content rather than merely reproducing the content produced by individuals during their lifetime. The authors focus on how to design such technologies while balancing their potential positive impacts with possible social and ethical risks.

Why is this issue important?

As generative AI becomes more widespread, its profound impact on cultural, personal, and social practices is inevitable. Specifically, it may transform traditional practices surrounding death, grief, and remembrance, and could even influence legal frameworks, estate planning, and religious beliefs. A comprehensive assessment of the far-reaching challenges and risks posed by this technology is crucial to ensuring positive societal development.

Research Motivation and Related Work

The authors draw on historical technological literature to reflect on the potential of AI in posthumous commemoration practices and review existing attempts, such as personal enthusiasts and startups using AI technology to create "generative ghosts." Furthermore, existing research related to digital legacies and posthumous practices (e.g., "digital inheritance" and "online memorials") highlights unresolved issues and challenges in integrating technology into cultural customs.

Solutions

What methods or solutions did the authors propose?

The authors propose a novel "design space framework" to describe the potential design dimensions of generative ghosts. These dimensions include: origin (voluntary vs. third-party-created ghosts), deployment timeline (pre-mortem vs. post-mortem), degree of anthropomorphism, patterns of representational evolution, embodiment, and more. They also analyze potential impacts and application scenarios from both technical and ethical perspectives.

What is innovative about this solution?

  • Introduced the new concept of "generative ghosts," expanding it into multifunctional systems capable of generating content and performing tasks, rather than merely reiterating data or content from the deceased.
  • Developed a multidimensional design space to systematically describe the characteristics of generative ghosts and how specific design choices might lead to particular risks or benefits.
  • Integrated technological advancements, cultural differences, and ethical norms to guide future research and practice on such technologies, aiming to avoid potential negative consequences.

What are the implementation steps? What key technologies were used?

The authors constructed the design space framework based on literature analysis and case studies, comparing the potential outcomes of different choices (e.g., interaction interface design, ghost maintenance methods). While specific implementation technologies were not detailed, they mentioned leveraging advancements in large language models (e.g., GPT-4, PaLM-2) and other generative AI technologies as the technical foundation.

Research Outcomes

What specific outcomes were achieved?

  1. Theoretical Contribution: Defined generative ghosts as a new socio-technical phenomenon and developed a structured design framework.
  2. Ethical and Social Analysis: Highlighted potential social and psychological impacts (e.g., aiding mourners or exacerbating complex grief), privacy risks, and economic implications of generative ghosts.
  3. Research Agenda: Proposed key areas for future research, such as user studies with diverse cultural groups, experimental designs based on relevant technologies, and the development of appropriate policies to mitigate potential risks.

What advantages does it have compared to existing solutions?

Compared to simple online memorials or existing low-complexity "griefbots," generative ghosts:

  • Leverage content generation capabilities to enable the deceased to "participate" in complex tasks or scenarios.
  • Incorporate more anthropomorphic and personalized features, enhancing interaction experiences and expanding the scope of application.
  • Offer more design possibilities, such as setting rules for ghost evolution or adding socio-economic capabilities to the ghost.

What are the experimental or evaluation results?

The paper does not provide direct experimental evaluation results; the authors' analysis is primarily based on literature and speculative frameworks. Suggested future work includes building prototypes for user studies and validating the impact of design choices.

Limitations and Future Directions

  • Limitations: The current research is primarily theoretical and framework-based, lacking concrete user data and implementation details; it does not yet address the needs and acceptance levels of all cultural groups.
  • Future Directions:
    • Further study the relationship between design dimensions and their specific impacts on individuals and society.
    • Explore the applicability and acceptance of generative ghosts across different cultures.
    • Develop prototype systems to evaluate actual user psychological and behavioral responses, particularly in relation to ethical considerations.
    • Design appropriate policies and legal frameworks to regulate the behavior of "third-party generative ghosts" and prevent misuse.

This research provides a valuable perspective for exploring the potential applications of AI in future societies and managing complex ethical and social issues, while also presenting a clear research agenda to advance the field.

Quick Actions

Share

Share this page

ios_share

https://hci.top/en/papers/chi/188425/2025

AdRecommended

Learn AI Coding at CodeNow

open_in_newOpen DOI Link
DOI: https://dl.acm.org/doi/10.1145/3706598.3713758
At a Glance

Paper Snapshot

fact_check
dataset
Source
CHI
calendar_month
Year
2025
emoji_events
Award
No award tagged
group
Authors
2 authors
sell
Subtopics
Generative AI (Text, Image, Music, Video), Algorithmic Fairness & Bias, Technology Ethics & Critical HCI
work
Professions
AI/ML Researchers & Engineers, HCI Researchers, Sociologists & Anthropologists
article
Content Status
Full text indexed
hub
Related Papers
3 related papers