The Digital Twin: Who Owns Your Story When an Algorithm Imitates You?

What the experiment showed, and what it did not

Researchers from Stanford and Google DeepMind held two-hour interviews with 1,052 Americans, then supplied language models with the transcripts so that they would answer General Social Survey questions in their owners' place. The simulation's accuracy came to 83% of "self-consistency", meaning the share of their own answers that people repeat after two weeks, rising to 86% when survey data were added. The benchmark relying on demographic traits alone stands at 74%.

This does not mean a human being has been cloned; the simulation here is of survey answers, in an American sample, in a laboratory. But it shows that a modest amount of conversation is enough to form a predictable picture.

When experience becomes raw material

Shoshana Zuboff, professor emerita at Harvard Business School, argues that part of the digital economy was built on treating human experience as a free raw material turned into behavioural data, a portion of which is kept as "behavioural surplus" that machines analyse and from which they produce predictions sold in markets the data's owner cannot see. This is her reading and it is contested, but it sums up a legitimate worry: that a picture of you is built in whose drawing you took no part.

Identity is a story, not a list of traits

The philosopher Paul Ricoeur distinguishes between "idem" identity, unchanging sameness, and "ipse" identity, the self across change. As the Stanford Encyclopedia of Philosophy summarises, we are neither what we were ten years ago nor someone else. Our identity is told, and attested by our own testimony and that of others.

Algorithmic prediction assumes the past determines tomorrow. A human being, though, repents, retreats, apologises and begins again. So a person should not be judged by an image inferred from him instead of being asked about it, nor denied the right to be other than he was.

In 2007 Nick Yee and Jeremy Bailenson described the "Proteus effect": in two experiments, people with more attractive avatars behaved with more openness and closeness to others, and people with taller avatars acted with greater confidence in negotiation. That is, our representation may influence our behaviour.

Whether this effect carries over to digital twins that speak in our names is a reasonable possibility whose details this study has not yet tested, so it is treated as a hypothesis.

After death

Two Cambridge researchers (2024) studied "griefbots" that simulate the deceased. They warned of giving people the illusion that the dead person is alive, of their effect on children, of the memory of the departed turning into a platform for covert advertising, and of the absence of the deceased's consent. They recommended clear disclosure that the speaker is a machine, age restrictions, and procedures for retiring these copies. The dead have a sanctity in our Sharia, and the ruling on this practice is referred to people of fatwa.

Your rights, and what to do

Saudi Arabia's Personal Data Protection Law, which organisations have had to comply with since 14 September 2024 and which the Saudi Data and Artificial Intelligence Authority oversees, gives data subjects rights, among them to be informed of processing, to access their data and obtain a copy, to correct it, to request destruction of what is no longer needed, and to lodge complaints.

In practice:

• Look at what platforms collect about you, and use your right to correction and destruction.
• Ask to be told when you are talking to a person and when to an automated agent.
• Do not authorise an agent to negotiate and sign in your name without knowing the limits of what you have delegated.
• Teach young people that their worth is not in the number of likes.

A human being is greater than any model that imitates him: he has the capacity to repent, to review and to rebuild, which no algorithm measures.