RetroChat: Designing for the Preservation of Past Digital Experiences

TL;DR

This paper proposes an interactive, AI-driven preservation approach to reclaim past digital experiences on Chinese social networks (2000–2010) by deploying RetroChat, a GPT-based agent that emulates era-appropriate language within a retro MSN-like interface. It builds a historical dialogue corpus from Tianya Club and Wayback Machine archives, and uses meticulous prompt engineering to instantiate a period-authentic netizen persona and conversational style. A qualitative study with 18 participants demonstrates that interacting with RetroChat elicits memory recall and nostalgia, with users sometimes adapting their language to match the past style and even revealing broader life memories. The work argues for digital heritage preservation as an experiential process—extending beyond static archives to enable first-person re-experiencing of historical online culture and informing future preservation methods.

Abstract

Rapid changes in social networks have transformed the way people express themselves, turning past neologisms, values, and mindsets embedded in these expressions into online heritage. How can we preserve these expressions as cultural heritage? Instead of traditional archiving methods for static material, we designed an interactive and experiential form of archiving for Chinese social networks. Using dialogue data from 2000-2010 on early Chinese social media, we developed a GPT-driven agent within a retro chat interface, emulating the language and expression style of the period for interaction. Results from a qualitative study with 18 participants show that the design captures the past chatting experience and evokes memory flashbacks and nostalgia feeling through conversation. Participants, particularly those familiar with the era, adapted their language to match the agent's chatting style. This study explores how the design of preservation methods for digital experiences can be informed by experiential representations supported by generative tools.

Figures (7)

• Figure 1: The figure presents user interfaces from early Chinese social networking platforms. On the left is a screenshot of the Tianya forum; the top middle displays the web-based Sina chatroom; the top right shows the 2005 version of QQ; the bottom middle features MSN Messenger 7.5; and the bottom right depicts the chat interface of Xiaonei on phone.
• Figure 2: The diagram indicates the sequence of our design, starting from the initial gathering of raw data, followed by corpus building based on the dataset, prompt engineering, and finally deploying RetroChat through an MSN service to enable interaction.
• Figure 3: From left to right, the results show the crawler data for Tianya Club, Sina BBS, and MOP, spanning the years 1999 to 2011. Below is an example illustrating the ammount and accessibility in 2005 for each BBS. Blue indicates that the web server's response code for the corresponding capture was successful, while green signifies that the crawler encountered a redirect status.
• Figure 4: The left figure shows the overall setup of the environment, the middle figure illustrates how users engage and interact with the artifact, and the right figure displays the legacy MSN 8.1 interface where participants chat with the agent.
• Figure 5: Experimental Procedure: After obtaining consent, the study is conducted in three stages. The first stage is the Pre-task, which involves collecting participant information, including their familiarity with Chinese SNS from 2000-2010. The second stage is the Chatting Task, where participants engage with RetroChat. The final stage is the Post-task, a retrospective interview featuring open-ended questions.