What if AI could spot your lies? | Riccardo Loconte - TED Talks Daily

Summary

Summary of "What if AI could spot your lies?" | Riccardo Loconte on TED Talks Daily

Introduction

In the February 17, 2025 episode of TED Talks Daily, hosted by Elise Hu, neuroscientist and psychologist Ricardo Leconte presents a compelling exploration into the intersection of artificial intelligence (AI) and lie detection. Leconte delves into the current limitations of human lie detection and examines how AI can potentially surpass human capabilities in identifying deception.

The Prevalence and Challenge of Lie Detection

Leconte begins by addressing the ubiquity of lying in daily human interactions. He states, "lying is very common and it is now well established that we lie on a daily basis" (02:44). Despite the frequency of deception, humans notoriously struggle to detect lies accurately. Leconte highlights that studies have shown humans lie approximately two times per day, though these figures are estimates.

He underscores the paradox that even experts in various fields, such as law enforcement and psychology, are only marginally better than chance at detecting deception. "naive judges accuracy was on average around 54%. Experts performed only slightly better with an accuracy rate around 55%," he notes (03:40). This revelation sets the stage for exploring how AI might bridge this significant gap.

AI and Lie Detection: The Potential and the Process

Transitioning to the core of his presentation, Leconte introduces the concept of leveraging AI, specifically large language models, to enhance lie detection. He explains, "AI could be used to detect lies. And you will be very surprised by the answer," indicating the promising yet complex nature of this technological advancement (04:30).

To illuminate his research approach, Leconte describes the process of fine-tuning large language models. He likens these models to students who receive specialized training beyond their general education. "fine tuning is that extra education," he metaphorically explains (05:15). This fine-tuning involves training AI on datasets comprising truthful and deceptive statements across various contexts, such as personal opinions, autobiographical memories, and future intentions.

Methodology and Experiments

Leconte outlines three primary experiments conducted to assess the efficacy of AI in lie detection:

Single Dataset Training: The AI model, Flanty 5 developed by Google, was fine-tuned on individual datasets. This approach yielded promising results, with the model achieving an accuracy range between 70% and 80% (12:10).
Cross-Dataset Training: The model was trained on pairs of datasets and tested on a third, unseen dataset. Here, Flanty 5's accuracy plummeted to nearly 50%, indicating a significant drop in performance (14:20). This decline suggested that while the AI performed well within familiar contexts, it struggled to generalize across different types of deceptive scenarios.
Combined Dataset Training: By integrating all three datasets into a larger training set, the model's accuracy rebounded to approximately 80% (16:05). This outcome demonstrated that comprehensive exposure during training allowed the AI to better generalize and maintain high accuracy across varied contexts.

Implications of AI-Enhanced Lie Detection

Leconte envisions a future where AI-driven lie detection is seamlessly integrated into numerous aspects of society. He paints a picture of enhanced security measures, more transparent political discourse, and improved hiring processes. "From tomorrow we can say when a politician is actually saying one thing and truly believe something else," he muses (17:00). Additionally, he anticipates applications in identifying malicious intentions during security screenings and reducing the prevalence of social media scams.

Ethical Considerations and Risks

Despite the optimistic outlook, Leconte cautions against uncritical reliance on AI for lie detection. He warns, "people will just be more likely to accuse others of lying just because an AI says so" (17:50). This blind trust could erode fundamental societal values like trust and critical thinking.

Leconte advocates for interpretability in AI systems, where the technology not only identifies deception but also provides transparent explanations for its judgments. "Imagine a world where AI doesn't just offer conclusions, but also provide clear and understandable explanations behind its decisions," he suggests (18:30). This approach would empower users to make informed decisions rather than passively accepting AI's assessments.

Conclusion

Ricardo Leconte's talk offers a nuanced perspective on the capabilities and challenges of using AI for lie detection. While AI holds the promise of significantly improving our ability to identify deception, Leconte emphasizes the necessity of ethical considerations and the preservation of human critical thinking. He envisions a future where AI serves as an augmentative tool, enhancing rather than replacing human judgment, ultimately fostering a more truthful and transparent society.

Notable Quotes

"lying is very common and it is now well established that we lie on a daily basis." — Ricardo Leconte (02:44)
"naive judges accuracy was on average around 54%. Experts performed only slightly better with an accuracy rate around 55%." — Ricardo Leconte (03:40)
"AI could be used to detect lies. And you will be very surprised by the answer." — Ricardo Leconte (04:30)
"fine tuning is that extra education." — Ricardo Leconte (05:15)
"From tomorrow we can say when a politician is actually saying one thing and truly believe something else." — Ricardo Leconte (17:00)
"people will just be more likely to accuse others of lying just because an AI says so." — Ricardo Leconte (17:50)
"Imagine a world where AI doesn't just offer conclusions, but also provide clear and understandable explanations behind its decisions." — Ricardo Leconte (18:30)

Final Thoughts

Ricardo Leconte’s insights illuminate both the transformative potential and the ethical dilemmas posed by AI in the realm of lie detection. As society stands on the brink of integrating such technology, his call for balanced, transparent, and interpretative AI systems serves as a crucial guide for responsible innovation.

Transcript

Boarding Announcement (0:01)

Ladies and gentlemen, we are now boarding Group A, please have your boarding passes ready to scan. If your phone is cracked old or was chewed up by your Chihuahua travel companion, please refrain from holding up the line and instead simply go to Verizon and trade in any phone in any condition from one of their top brands for the new Samsung Galaxy S25 plus with Galaxy AI on Unlimited ultimate and a watch or tab. Also on app Service plan required for watch or tab. Trade in and additional terms apply. See verizon.com for details.

Freddy Wong (0:30)

Hi, this is Freddy Wong from Dungeons and Daddies, and this episode is sponsored by Rocket Money. Houston. Houston, we have a problem, and that's too many subscriptions that I don't know about because I like to put my credit card number into sites just for the sheer thrill of it. That's the fundamental problem of the Internet and money, and Rocket Money is here to solve that. Rocket Money is a personal finance app that helps find and cancel your unwanted subscriptions, monitors your spending, and helps lower your bills. You can see all those subscriptions that you've accrued over a lifetime of putting your credit card in on the Internet in one place. If you don't want them, just cancel them with a few taps. Rocket Money can help with that. Rocket Money has over 5 million users and has saved a total of $500 million in canceled subscriptions, saving members up to $740 a year when using all the app's premium features. Stop wasting money on things you don't use. Cancel your unwanted subscriptions by going to rocketmoney.com cancelsubs. That's rocketmoney.com cancelsubs, not submarines.

Elise Hu (1:32)

TED Talks Daily is sponsored by Capital One. In my house, we subscribe to everything. Music, TV, even dog food. And it rocks. Until you have to manage it all, which is where Capital One comes in. Capital One credit card holders can easily track, block or cancel recurring charges right from the Capital One mobile app at no additional cost. With one sign in, you can manage all your subscriptions all in one place. Learn more@Capital1.com Subscriptions Terms and Conditions apply. You're listening to TED Talks Daily, where we bring you new ideas to spark your curiosity every day. I'm your host, Elise Hu. Neuroscience researcher Ricardo Laconte specializes in something that is relevant across many, many fields. How to detect lies or lying. In his 2024 talk, he sheds light on how AI can be assistive and in some cases, better than humans at lie detection. But he warns there are also areas where AI has a ways to go.

Ricardo Leconte (2:44)

This is something you won't like, but here everyone is a liar. Don't take it too personally. What I mean is that lying is very common and it is now well established that we lie on a daily basis. Indeed, scientists have estimated that we tell around two lies per day. Although of course it's not that easy to establish those numbers with certainty. And well, I introduce myself. I'm Ricardo. I'm a psychologist and a Ph.D. ph.D. Candidate. And for my research project I study how good are people at detecting lies. Seems cool, right? But I'm not joking. And you might wonder why a psychologist was then invited to give a TED talk about AI. Well, I'm here today because I'm about to tell you how AI could be used to detect lies. And you will be very surprised by the answer. But first of all, when is it relevant to detect lies? A first clear example that comes to my mind is in the criminal investigation field. Imagine you are a police officer and you want to interview a suspect and the suspect is providing some information to you and those information are actually leading to the next steps of the investigation. We certainly want to understand if this suspect is reliable or if they are trying to deceive us. Then another example comes to my mind and I think this really affects all of us. So please raise your hands if you would like to know if your partner cheated on you. And don't be shy because I know you see, it's very relevant. However, I have to say that we as humans are very bad at detecting lies. In fact, many studies have already confirmed that when people are asked to judge if someone is lying or not, without knowing much about that person or the context, people's accuracy is no better than the chance level. About the same as flipping a coin. You might also wonder if experts such as police officers, prosecutors, experts and even psychologists are better at detecting lies. And the answer is complex because experience alone doesn't seem to be enough to help detecting lies accurately. It might help, but it's not enough to give you some numbers. In a well known meta analysis that previous scholars did in 2006, they found that naive judges accuracy was on average around 54%. Experts performed only slightly better with an accuracy rate around 55%. Not that impressive, right? And those numbers actually come from the analysis of the results or 108 studies. Meaning that these findings are quite robust. And of course the debate is also much more complicated than this and also more nuanced. But here the main take home message is that humans are not good at detecting lies. What if we are creating an AI tool where everyone can detect if someone else is lying. This is not possible yet, so please don't panic. But this is what we tried to do in a recent study that I did together with my brilliant colleagues, whom I need to thank. And actually, to let you understand what we did in our study, I need to first introduce you to some technical concepts and to the main characters of this story. Large language models. Large language models are AI systems designed to generate outputs in natural language in a way that almost mimics human communication. If you are wondering how we teach these AI systems to detect lies, here is where something called fine tuning comes in. But let's use a metaphor. Imagine large language models being a student who have gone through years of school learning a little bit about everything, such as language concepts, facts. But when it's time for them to specialize, like in law school or in medical school, they need more focused training. Fine tuning is that extra education. And of course, large language models don't learn as humans do. But this is just to give you the main idea. Then, as for training students, you need books, lectures, examples. For training large language models, you need data sets. And for our study, we consider three data sets. One about personal opinions, one about past autobiographical memories, and one about future intentions. And these data sets were already available from previous studies and contained both truthful and deceptive statements. Typically, you collect these types of statements by asking participants to tell the truth or to lie about something. For example, if I was a participant in the truthful condition and the task was tell me about your past holidays, then I would tell the researcher about my previous holidays in Vietnam. For the deceptive condition, they will randomly pick some of you who have never been to Vietnam, and they will ask you to make up a story and convince someone else that you've really been to Vietnam. And this is how it typically works. And as in all university courses, you might know this after lectures, you have exams. And likewise, after training our AI models, we would like to test them. And the procedure that we followed, that is actually the typical one, is the following. So we picked some statements randomly from each data set and we took them apart. So the model never saw these statements during the training phase and only after the training was completed. We used them as a test, as the final exam. But who was our student then? In this case, it was a large language model developed by Google and called Flanty 5 Flanny for friends. And now that we have all the pieces of the process together, we can actually dig deep into our study. Our study was composed by three main experiments. For the first experiment, we fine tuned our model, our Flanty 5 on each single dataset separately. For the second experiment, we fine tuned our model on two pairs of data set together and we test it on the third remaining one and we use all three possible combinations. For the last final experiment, we fine tuned the model on a new larger training test set that we obtained by combining all the three datasets together. The results were quite interesting because what we found was that in the first experiment, Flanty 5 achieved an accuracy range between 70 and 80%. However, in the second experiment, Flanty 5 dropped its accuracy to almost 50%. And then surprisingly, in the third experiment, Flanty 5 rose back to almost 80%. But what does this mean? What can we learn from these Results? From Experiment 1 and 3, we learned that language models can effectively classify statements as deceptive, outperforming human benchmarks and aligning with previous machine learning and deep learning models that previous studies trained on the same data set. However, from the second experiment we see that language models struggle in generalizing this knowledge, this learning across different contexts. And this is apparently because there is no one single universal rule of deception that we can easily apply in every context. But linguistic use of deception are context dependent. And from the third experiment we learned that actually language models can generalize well across different contexts if only they have been previously exposed to examples during the training phase. And I think this sounds as good news. But while this means that language models can be effectively applied for real life applications in light detection, more application is needed because a single study is never enough so that from tomorrow we can all have these AI systems on our smartphones and start detecting other people's lives. But as a scientist, I have a vivid imagination and I would like to dream big. And also I would like to bring you with me this futuristic journey for a while. So please imagine me living in a world where this lie detection technology is well integrated in our life, making everything from national security to social media a little bit safer. And imagine having this AI system that could actually spot fake opinions. From tomorrow we could say when a politician is actually saying one thing and truly believe something else. And what about the security board context where people are asked about their intentions and reasons for why they are crossing borders or boarding planes? Well, with these systems we could actually spot malicious intentions before they even happen. And what about the recruiting process? We heard about this already, but actually companies could employ this AI to distinguish those who are really passionate about the role from those who are just trying to say the right things to get the job. And finally we have social media scammers trying to deceive you or to steal your identity. All gone. And someone else may claim something about fake news and well, perfectly let language model could automatically read the news, flag them as deceptive or faked and we could even provide users with a credibility score for the information they read. It sounds like a brilliant future, right? Yes, but all great progress comes with risks. As much as I'm excited about this future, I think we need to be careful. If we are not cautious, in my view, we could end up in a world where people might just blindly believe AI outputs. And I'm afraid this means that people will just be more likely to accuse others of lying just because an AI says so. And I'm not the only one with this view because another study already proved it. In addition, if we totally rely on this lie detection technology to say someone else is lying or not, we risk losing another important key value in society. We lose trust. We won't need to trust people anymore because what we will do is just asking an AI to double check for us. But are we really willing to blindly believe AI and give up our critical thinking? I think that's the future we need to avoid. What I hope for the future is more interpretability. And I'm about to tell you what I mean. Similar to when we look at Travis online and we can both look at the the total number of stars place us, but also we can look at more in detail at the positive and negative reviews and try to understand what were the positive sides, but also what might have gone wrong to eventually create our own and personal idea if that is the place where we want to go, where we want to be. Likewise, imagine a world where AI doesn't just offer conclusions, but but also provide clear and understandable explanations behind its decisions. And I envision a future where this lie detection technology wouldn't just provide us with a simple judgment, but also with clear explanations for why it thinks someone else is lying. And I would like a future where yes, this light detection technology is integrated in our life or also AI technology in general. But still at the same time we are able to think critically and decide when we want to trust any judgment or when we want to question it. To conclude, I think the future of using AI for light detection is not just about technological advancements, but about enhancing our understanding and fostering trust. It's about developing tools that don't replace human judgments, but empower it, ensuring that we remain at the helm. Don't step into a future with blind reliance on technology. Let's commit to deep understanding and ethical use, and we pursue the truth. Thank you.