MICHEL MARTIN, HOST:
Now that we've heard about the broad threats of disinformation, let's look at a specific tool often used to spread disinformation or misinformation - deep fakes. This is when technology is used to alter an image or video of a person in a manner that's hard to detect with the naked eye, often to embarrass someone. Earlier this month, that technology was put to use in India, but it wasn't used by an opponent or enemy. It was used as a campaign tool.
The day before legislative assembly elections in Delhi, the president of the ruling party, Manoj Tiwari, released two videos criticising his opponent. In one video, Tiwari speaks English, and in the other he speaks, in a Hindi dialect. The videos went viral on WhatsApp. The only problem is those videos were deepfakes. to worry. Tiwari doesn't speak either English or that dialect, a fact uncovered by a reporter for Vice.
We wanted to talk more about the implications of this, so we've called Tarunima Prabhakar. She is the co-founder of a project to address misinformation on chat apps and encrypted networks. And she is with us now from Pune, India. Tarunima Prabhakar, thanks so much for joining us.
TARUNIMA PRABHAKAR: Thank you so much, Michel.
MARTIN: Now, when we've talked about deepfakes in the U.S. context, we're mainly talking about videos altered by politicians to make their opponents look bad or altered by politicians or people who like those people or want to support those people to make somebody opposing them look bad. But that's not what happened here. The candidates' own people did this, right? So do you see that as a positive use for deepfakes or is there no such thing?
PRABHAKAR: Well, so I think it's interesting that, you know, the party and also the marketing company behind it pitched it as a positive campaign. But I think it's important to put this in context of the Delhi electorate, which is the Delhi is a very - has very rich ethnic and linguistic diversity. And so speaking in a language is a way of establishing a connection with the demographic. It's a way for the candidate to say, hey, I speak your language. I'm one of you. All political candidates, whether it's in India or in the United States, are trying to appeal more relatable to their voters.
And so, you know, if the aim was to just get the message out to people in more languages, there were other means of doing it. So you could have dubbed the video. One could have put subtitles to it. But when someone is creating a deepfake and trying to imply that they speak a language that they're not, in some way, they're trying to appeal more relatable on a trait that they do not possess.
MARTIN: They're basically lying about who they are. They're lying about a skill that they actually don't have.
PRABHAKAR: Yes, in a small way, they are, yes.
MARTIN: Could you just talk a little bit more about about why you think the public needs to be cautious about this, I mean, why you think it is that people need to to really think about the ethics of this?
PRABHAKAR: So I think - I treat these as two separate questions. Why does the public need to be cautious about this is because, at least in India, you know, a lot of this content is circulating on encrypted chat apps. And so for content that is being circulated on Facebook and Twitter, there is centralized content moderation, so that the platform has a lot of control in taking down let's say deepfakes or other problematic content. But when you have messages and content being circulated on chat apps such as WhatsApp, there is no centralized authority that can take down content. And so in some ways the resiliency for - against disinformation and misinformation comes via news literacy and comes via sort of like over news consumption.
And the second question that you asked was a question about ethics. And so I think there can be a case for, you know, deep learning or AI-generated videos that can be useful in a certain context. But that's a very broad discussion. I think it's useful to narrow to a specific discussion of election campaigning. And the question is, can deepfakes be useful in any regard in election campaigning? And some people, especially after this Vice article, have claimed that perhaps not because authenticity is very important, authenticity about the candidates. So even if a candidate is producing videos about themselves and they're endorsing it, it's still in some ways inauthentic.
MARTIN: Well, before I let you go, is there anything that the public can do to avoid being taken in? I mean, this is a situation where sharp-eyed observers who were following the campaign knew what the truth was and called the candidate out on it. So is there anything that the public can do to better educate itself or to avoid being taken in?
PRABHAKAR: Sure. So I think in the U.S. context as well, people have started, you know, a lot of researchers are working on cues that individuals can use to sort of investigate videos. So often the way eyes blink is a good indication of, you know, whether it's a deepfake or not. In this case, the way the lips were moving, it was definitely a little odd. And so one has to be a very, very careful consultant consumer and sort of has to watch these videos very carefully. It's not that important to know that the video has been generated using deepfakes as it is important to know that it is fake and this is not a true video and the candidate does not speak in this language.
MARTIN: That was Tarunima Prabhakar. She is the co-founder of Tattle. That's a civic tech project that aims to make verified information more accessible to mobile users in India. Tarunima, thank you so much for talking to us. I hope we'll talk again.
PRABHAKAR: Thank you, Michel. Transcript provided by NPR, Copyright NPR.