Deepfakes - minefields and opportunities

Deepfakes are a minefield. They’re both terrifying (swaying election results) and offer opportunity (malaria no more).

So what is a deepfake?

The Oxford Dictionary describes a deepfake as:

“a video of a person in which their face or body has been digitally altered so that they appear to be someone else, typically used maliciously or to spread false information.”

One notable thing is that a deepfake is not always used maliciously and we can use it positively. A question that springs to my mind is:

Can we use a deepfake as a tool to achieve honest feedback?

Imagine you’re a senior leader who wants honest employee feedback about their working environment. What do they genuinely think??

Questionnaires only go so far. One-to-one dialogues uncover much more. As a senior leader, approaching people directly and encouraging honest feedback is a good start but they’ll need psychological safety.?

You’re senior to them. You have the power. Will people feel safe enough to give you honest feedback?

What if they can become deepfakes?

By digitally altering their voice and appearance in real-time, will they give more honest feedback??

In answering this question, I’m making some assumptions:

• There is no way anyone can access information that identifies who is giving me feedback, and the contributor knows this.
• If I accidentally figure out who the contributor is, I’ll keep this in confidence from everyone, including them.
• They genuinely want to give honest feedback.
• I can make sense of what they’re saying without knowing their background.

With that in mind I could do an experiment to assess the impact of using deepfakes BUT how do I measure success??

How do I know that people have given more honest feedback than they would have otherwise??

How do I measure that it’s made a difference?

I could ask contributors whether they gave more honest feedback as a result of the deepfake. This still relies on my assumptions being true. I could run some feedback sessions without deepfakes and compare the feedback. However, this is likely to have a strong dependency on the individuals selected.

So in answer to my question:

Can we use a deepfake as a tool to achieve honest feedback?

The jury’s out. I’m not convinced. I’m genuinely interested:

What do you think?