TL;DR
What We're Asking

You're invited to participate in a one-week online experiment (November 9-16) exploring how people learn history through AI conversations.

What you'll do:

  • Have conversations with an AI trained on Dutch-Chilean solidarity stories

  • Use either email or chat (we'll assign you randomly)

  • Interact at least 3 times during the week (we'll send reminders)

  • Complete a brief survey at the end (8-12 minutes)

What you'll learn:

  • Stories from Chilean refugees and Dutch solidarity activists from the 1970s-80s

  • How different communication speeds affect your engagement with historical content.

Time commitment: Flexible and self-paced. Each conversation can be as short or long as you want. At the end of the study, we ask you to complete a brief form (8-12 minutes). In total, we estimate that your participation would take between 1 and 1.5 hours of your time.

Requirements: None! No prior historical knowledge is required, and no special software is needed (except for an email or web browser); just a curiosity about history and technology.

Before starting, you'll be asked to sign a consent form and will then receive an email with instructions on how to begin.

Introduction

How does the way we interact with AI influence our understanding of history?

A growing body of research in heritage and media studies shows that our memory of the past is not a fixed record of facts—it's a dynamic process of reconstructing history through the lens of the present. When we think about historical events, we often recall them with different emphases, blending facts and stories with today's questions, emotions, and contexts. In this way, we actively shape our understanding of the past through dialogue with our present.

Artificial intelligence, particularly language models, is increasingly becoming part of this dialogue. In museums and educational apps, we are now having the ability to "speak" with the past—whether through chat agents retrieving information from historical documents or through interactions with AI personas representing historical periods. These exchanges don't just deliver facts—they shape how we understand and feel about history: what gets emphasised, how we think about the information, and our engagement with historical narratives.

This research asks: What if we slow down the conversation? What if, instead of defaulting to synchronous, real-time chat, we communicate with language models via email? Does the interface we use to interact with AI change how thoughtfully we engage with historical content and how we critically assess what we're learning?

This experiment focuses on the Dutch-Chilean solidarity movement of the 1970s and 1980s—a period when many Chileans fled to the Netherlands during the military dictatorship and were supported by Dutch solidarity networks. You don't need any prior knowledge about this historical period; the goal is to learn about these communities through your interactions with the AI system.

This exploratory research does not aim for statistically significant results. Instead, it focuses on opening questions about how different ways of communicating with AI might influence our learning experience and critical engagement with historical content.

The Experiment

To explore how different communication modes affect engagement with AI-presented historical content, we have fine-tuned a language model using testimonies from individuals involved in the Dutch-Chilean solidarity movement of the 1970s–1980s. These testimonies were collected in the book 40 jaar, 40 verhalen: Chileense vluchtelingen en solidariteit, edited by historian Jan de Kievid and published in 2013.

The language model will present fictional characters based on these real testimonies. You are being informed now about this fictional mediation—during the actual experiment, you'll simply engage in conversation with these AI personas to learn about their historical experiences.

This fully online experiment will involve around ten participants from different backgrounds. Some participants will engage with the system via immediate chat, others via email correspondence.

The experiment runs from Sunday 9 November to Sunday 16 November. You will receive information on the starting day, but the experiment is self-paced—you don't need to begin immediately.

What's required from you

  • Engage with the language model (either via email or chat, depending on the assigned platform) over a period of one week, with a minimum of three sessions. It will be up to you to decide how long the sessions may take. They can range from one minute to one hour or more.

  • You'll receive reminders to help you stay engaged.

  • Complete a survey at the end of the study (8-12 minutes)

Presentation and Results

The results of this experiment will be presented at the GenAI & Creative Practices conference, taking place on December 17 and 18 at the University of Amsterdam. Unfortunately, this event is not open to the public, but I'll be happy to share the key insights with you via email during January once the analysis is complete.

Would You Like to Be Part of This Research?

Since you're reading this either through direct contact with me or because someone I know recommended you for this research, I hope this website has provided you with a clearer understanding of the "Time to Think" project. If you're interested in participating, please reach out to me directly—whether by email, text message, or any other method that works for you.

Before joining the project, you will be asked to read a brief information brochure that summarises what's presented on this website, and you'll need to sign a consent form regarding the use of your data. All data collected will be anonymised and stored securely on servers connected to the University of Amsterdam, in accordance with standard academic research protocols.

I'm happy to answer any questions you might have about the process, timing, or what to expect.

Looking forward to hearing from you.

Pablo Núñez Palma
PhD Candidate
University of Amsterdam