Digital twin tech is a double-edged sword

A digital twin is a simulation of an object, system, process, or person.

The ability to make and use digital twins is one of the many new opportunities for business that didn’t exist for most businesses before the recent LLM-based AI revolution. I detailed in this space early last year how detailed twins of systems, such as factories or airplanes, can boost efficiency and save lives. 

That was focused on factories and processes, but we’ve been hearing a lot lately about digital twin replicas of actual, individual people. And while there are real potential benefits to the idea, it turns out there are some drawbacks as well. 

First, the benefits. 

Why we should double down on digital twins

One promising area for digital twins of specific people is in the realm of personalized medicine. Normally, medical interventions are one-size-fits-all, even though human biology varies greatly from one person to the next. 

Digital twin tech is already showing potential in improving individualized diabetes management. With a digital twin AI, doctors could create a dynamic, real-time virtual representation of a patient. They can then safely experiment with adjustments to treatment plans, including medication, diet, and lifestyle recommendations on the twin to find the best outcome. 

Once the treatment of the twin is optimized, it can be offered to the real person. 

Beyond diabetes specifically, this general approach could solve problems currently suffered by patients, such as the overprescription of medications, addiction to painkillers, and harmful reactions to various drugs. 

Digital twin tech is also seen as a way to optimize what is called “P4 medicine” (Predictive, Preventive, Personalized, and Participatory). 

Research tools

Digital twin tools are being used by scientists as research tools because it’s much easier, faster and cheaper to use digital twins than actual people. 

MIT and Oak Ridge wanted to figure out exactly which jobs could potentially be automated in the near future so that people and companies linked to those jobs can retrain. So, as part of their “Project Iceberg,” they made digital copies of every US worker — 151 million across 32,000 skills and 923 occupations. 

They concluded from the research that some 11.7% of jobs could now be automated with current technology (representing about 21 million workers and 1.2 trillion dollars in wages).

But it’s not at all clear that digital twins provide accurate results in research. 

Researchers from Columbia, Barnard, Yale, and Yeshiva compared survey responses of around 500 people (who had actually taken the survey) with digital copies of each of the participants looking at multiple baselines across 164 outcomes, spanning politics, social preferences, cognition, tech use, and other matters. 

The digital twins reproduced individuals’ responses with about 75% accuracy, a level of accuracy no better than generic demographic personas. 

Representing opinions and knowledge

Silicon Valley companies have emerged that enable you to replicate yourself. These include Viven, Synthesia, Touchcast, Delphi, and HeyGen.

Some experts predict that everyone will soon have their own digital twin. 

For now, the first movers tend to be Silicon Valley executives, including Eric Yuan (Zoom CEO), Sam Liang (Otter.ai CEO), Dan Thomson (Sensay CEO), Kevin Davis (Persona Studios CEO), and Garg (Viven CEO). These busy leaders are trying to outsource some of their work to their twins. As a great example, Delphi has some of their executives’digital twins available for chat on their home page, as well as a few other motivational people — even Arnold Schwarzenegger. 

Other prominent people with huge public demand for their time and advice are also offering digital twins of themselves: 

Jack Nicklaus – “Digital Jack” uses an autonomously animated avatar to interact with golf fans at scale, positioned as a way to share his expertise and brand.

Carmelo Anthony – His twin is meant to connect with fans and commercial partners, effectively multiplying his availability for branded interactions.

Mark Tuan — “Digital Mark” is the name of the K-pop star’s avatar, which answers fan questions instantly and can talk with thousands of fans at the same time.

And while some prominent people are creating twins of themselves to interact with their customers and fans, some companies are making digital twins of their own customers and fans. 

Creating digital twins of a company’s customers involves building real-time virtual models that represent individual customers. Trained on actual customer data, the twins are used to gain insights into consumer behavior, preferences, and needs without involving the real people. 

This can be helpful, but tricky. 

A new study by First Insight found that nearly 70% of US shoppers say they would trust brands less if those brands replaced real customer feedback with “digital twins” and synthetic personas. The study, based on responses from 1,303 adults ages 18 and up, found that 48% had never heard the term “digital twin” before it was explained as a digital replica built from purchases, browsing and other data. Once they understood it, 69% said they would lose trust if a brand relied on digital twins instead of asking customers directly.

Only 8% said they would prefer brands to simulate their preferences with AI rather than seek direct feedback, while 55% favored brands that simply ask them what they want. 

More than half (58%) said they would become detractors and might warn others if they learned a brand was using digital replicas or synthetic personas instead of genuine feedback, turning a back-end analytics choice into a word-of-mouth risk.

The study shows that the public doesn’t trust the idea of being represented by a digital twin. 

We have other data points from the world of social networking. 

Meta has added several AI-driven, public-facing “digital twin” features to its social apps, and the response has been mostly negative. On Instagram and Facebook, it tested AI character profiles that looked and behaved like regular users, posting images and replying to comments and DMs as if they were real people. 

Meta also rolled out tools for creators that let them generate video clips in which an AI version of their own face and voice reads scripted lines, so a digital stand-in can appear in Reels or Stories without the creator recording each take. This is effectively a social video twin aimed at branded content and fan engagement. 

After user complaints and critical news coverage, Meta removed or hid many of the AI character accounts and scaled back the feature set, a sign that the public is wary of unlabeled or hyper-realistic social media twins that can easily be mistaken for human beings.

Digital twins are a double-edged sword

Using digital twin technology to simulate real people can be powerful. But when it’s used to represent the views of people (leaders, celebrities, customers, or the public), people tend to oppose the idea. When they interact with prominent people, they want to interact with the real people. When their views are considered, they want to be asked rather than represented by a digital twin. 

This might change later, but for now the lesson is clear: it’s important to use digital twin tech carefully — and watch out for reputation harm from using it wrongly.