I’ve Always Been Jealous of Other People’s Families

“This project creates a lens to reflect on the human condition and the algorithms and systems we are increasingly relying on to interpret our world.”

Sarah Wolozin, Director, Open Documentary Lab, MIT


I’ve Always Been Jealous of Other People’s Families is an interactive installation that provides insight into the inner world of a type of machine learning system known as a generative adversarial network (GAN). The piece is a prologue to Marrow, a project that explores the possibility that intelligent machines could develop mental disorders.

In this project, four participants are asked to gather around a dinner table and act like a family. The table is set with sound, projection, and text, eliciting something familiar yet disturbing. Participants’ inputs are processed through the logic of GAN via varied machine learning networks trained on a data set of “perfect family dinners.” The output projected is a distortion, revealing an attempt to fill in the gaps between fantasy and reality. Looking at these outputs exposes hidden pockets of error and loss. Can this be the portal to the machine’s mental state?

This project was incubated at and supported by the MIT Open Documentary Lab.


Video © 2018 NFB. Credit: Annegien van Doorn.
I’ve Always Been Jealous of Other People’s Families, November 2018. IDFA DocLab.



Project lead:
shirin anlen, Open Documentary Fellow (2017-2019), MIT

Launched: The prologue of Marrow, I’ve Always Been Jealous of Other People’s Families, was presented at IDFA DocLab in November 2018.

On view: Visit the project website for news and information about how to experience I’ve Always Been Jealous of Other People’s Families


Avner Peled
Laura Juo-Hsin Chen
Christoval Valenzuela
Emma Dessau
Atlas V /producer Arnaud Colinart
National Film Board of Canada / producers Hugues Sweeney and Louis-Richard Temblay
MIT Open Documentary Lab
Raycaster, Ziv Schneider


Image courtesy of: shirin anlen.
Screen recording of a participant’s dataset demonstrates the DensePose estimations and image translation being done in real-time.
Screen recording of one person via “perfect family dinner” dataset. This shows the DensePose estimations and the image translation that is being done in real-time.


More information

Immerse: Family2Family: First Steps

International Documentary Film Festival Amsterdam (IDFA)

National Film Board of Canada



An exploration of how mental illness might emerge in intelligent machines