top of page
An AI-created image of a woman mixed with symbolism representing digital memory

About

Synthetic Pasts explores automated and algorithmic processes designed to change the structure and behaviour of media fragments from the past, in-so-doing, offering them new kinds of ‘afterlife’. These developments - for example, creating ‘holographic’ resurrections of celebrities, deepfake historical figures, animated photos from archives, or voice assistants powered with our deceased relatives’ voice data - connect viscerally with urgent discussions about what futures AI heralds.

 

While there have always been myriad and competing narratives about the past produced from personal and institutional archives, recent advances in deep learning technologies that change, manipulate, or create new media objects present unprecedented challenges. 

For the first time Synthetic Pasts asks what values and cultural assumptions are embedded in the systems that produce synthetic pasts, and what ethical challenges they present; for example, in relation to datafication, exploitation, bias, deception and (often posthumous) consent. In-so-doing it speaks to human concerns that reach beyond the digital, such as how we grieve, how we express and understand identity, how we articulate and commemorate the past, and the impacts of disinformation. In centring such ethical ambiguities,

 

Synthetic Pasts recognises the ways algorithms intervene within and are shaped by society and culture, including how they mediate our processes of memory and memorization.

An AI-created image of a man mixed with symbolism representing digital memory

Methodology

The project’s work packages together comprise an innovative quali-quant approach to synthetic pasts. Through those work packages we build on and expand approaches from different disciplines, combining and testing them to develop new insights at the intersection of Memory Studies, Critical Algorithm Studies and Digital Heritage.

 

Synthetic Pasts works creatively across multi-method empirical approaches, addressing current methodological limitations in the above disciplines, which are increasingly in need of new tools to make sense of these complex sites of study. We hope to pave the way for more methodologically curious studies in this nascent field of research.

Image credits: Google's Duet AI, 30 Nov 2023, prompts by JK_

VIGNETTE 1 

In February 2021 people shared short, animated videos featuring their deceased ancestors across social networks. These had been created with MyHeritage’s Deep Nostalgia application which uses blueprint (‘driver’) videos to bring images from the past ‘to life’; the deceased person is seen smiling, blinking, and turning their head. These videos became a runaway networked phenomenon as more than 30 million people animated photos and shared them with their followers. 

bottom of page