The Lives of Literary Characters

We're finished!

The Social Lives of Literary Characters project is now complete!

We want to extend our gratitude to all of you for your incredible contributions. Your participation played a vital role in the success of the first ever citizen science project related to understanding literature. We were overwhelmed by the enthusiasm and quality of your participation.

Thanks to your work, we have built a dataset consisting of over 73,000 annotations that helps illuminate the social lives of fictional characters. As promised, we share the data, our first AI model, and an open-access article describing our findings.

The key findings that were enabled by this research were:

1. Fictional social interactions are distinguished by their focus on embodied behavior (contact and observation). This supports an emerging body of work suggesting how fictional stories train our minds to think through our bodies and sensory experience.

2. Fictional social networks are far denser and less modular (have fewer sub-communities) than non-fictional networks. This suggests that one of the core values of fictional storytelling is to promote social cognition (to think through people who are interacting with others).

3. The identifcation of character interactions remains a challenging task! Our AI models still struggle to accurately rule out when two people are not interacting even if they are close to each other in a sentence. Your work helped show just how common it is for two people to NOT be interacting even when grammatically near each other.

4. A lightweight, open-access model trained on your data can outperform even the biggest closed frontier models like GPT-4!

Thank you again for your participation in this project. We hope you’ve enjoyed being part of this project and that you will continue to participate in future projects, which you can learn about at our Citizen Readers organization page.

Warm regards,
The Citizen Readers Research Team