Nemesysco: How 007 knows if you are lying

On the occasion of the 2nd Insur-Tech ecosystem meeting on Monday,  February 5th, which took place at the PwC premises in Zurich-Oerlikon 13 InsurTechs from the Israel and Swiss ecosystem presented their products and services. An insurtech called Nemesysco and its subsidiary Emotion Logic were of particular interest to thebroker.

Thebroker talks to Amir Liberman, CEO & Owner  Nemesysco Ltd & Emotion Logic Ltd.

Amir, you visited Switzerland during Insurtech Israel’s roadshow in Europe. Did you see anything at all of our country?

I love Switzerland, the wonderful people and the beautiful sights. It’s the 3^rd time I’m visiting it now, but no doubt I need to see much more of it…

How does the technology you invented back in 1997 work?

The technology core is using pretty simple principles: we capture the baseline behavior of different subtle vocal parameters that are sensitive to emotional changes, and we measure them to detect fluctuations during the conversation. These unique vocal properties were found through our long years of research to correlate with the most fundamental emotional states, and are not fully controlled by the speaker, yet they are so minute that using our naked ear we can hardly detect them.

It’s important to say that these vocal properties are the same for all mankind, and go beyond the age, ethnicity, race or gender. Our technology doesn’t care about language, words or culture and we’ve tested it literally all over the world just to ensure these vocal properties are completely objective and bias free.

After you invented it so long ago: How well can you read people without it?

I think I’m pretty good in feeling how others feel when I speak with them. I also think I have a good sense when people are not honest with me, but there were times when the technology and I had a different opinion about a matter… The technology was always correct and much more accurate than my own intuition.

Nemesysco’s core technology is called „Layered Voice Analysis“ (LVA). What exactly does it do?

It’s an excellent and multi-faced answer. At the beginning it was called MLVA (Multi layered voice analysis) but the name was too cumbersome. Layered Voice Analysis (LVA) means all the followings:

a. Essence: It represents the many depths and range of emotions the technology reaches and detect.

b. Technical: It describes the different processes we use to extract the emotional biomarkers. In a way, it is done layer after layer.

c. The results structure: The emotional output is also created and sorted in different layers based on the internal logic and emotions hierarchy.

Your voice analysis tool InTone.AI™ supports call centers. Also from insurance companies?

InTone.AI is a real-time set of solutions for call centers from Nemesysco (the parent company), and supports many different use-cases, from basic emotion made to ensure your clients satisfaction, to support your sales people phone approach and training, to underwriting and fraud detection, and there are even InTone versions for market research and medical uses. InTone.AI for fraud detection in the Insurance spaces comes with carefully crafted conversation scripts and requires some training to be used effectively. It is used by insurance companies today when the need is for real-time negotiation and immediate feedback. When there is no need for real time feedback, the new Emotion Logic platform offers some additional easy to use and automated solutions.

What data can be obtained with the InTone.ai Risk as a risk assessment system for insurance companies?

InTone.AI risk mode uses carefully created questionnaires and instructions to ensure both compliance with the regulations and laws on one hand, and optimize the input for voice analysis on the other. For example, once consent is obtained, the questions aim to be open-ended and avoid yes/no answers, so the voice analysis process can “listen” to the way the answer is created in the mind of the speaker. Honest responses are typically much easier to formulate and are easily spoken. When the system senses a stronger challenge to reply (stronger challenge compared to the other questions) the system will indicate it as higher risk, for secondary review by the managers. The system can also identify signs of excitement and stress, acting, hesitations and several more indications, all made to make a fraudulent claim easier to detect.

With the information you get, how do you know what a call partner thinks?

We have a very sophisticated technology, but we cannot do magic or read peoples’ mind. All we can do is report how they feel about what they just said. These indications, together with the operator’s human understanding can reveal more about the items that are relevant to the case and its investigation if it is indeed a high-risk case.

Does it mean that your tool can detect when a customer makes a claim and lies about it?

InTone.AI RISK for insurances and the tools available from the Emotion Logic hub identify fraudulent attempts, this is what they were made to do, but we refrain from looking at them “a lie catchers”. Our mission as I see it is to work within the scope of the law and legitimate needs under the GDPR to ensure first and foremost, effective and fair treatment to the LEGITIMATE claims.

Most of the clients are honest and deserve fair and fast processing of their claims at their time of need. In contrast, comes the insurance company’s requirement to prevent fraud, fight money laundering, and reduce the premium so everyone can purchase insurance. Of course, when our solutions warn about high-risk case, it must be investigated by additional means to prove the case, but now the investigation is focused on the few that are most likely to be found fraudulent, and with good guiding hints.

„I believe most people are decent, and I know sometimes people lie to survive. I know that few will cheat because they choose so, to live off the backs of others. I do not accept anyone should tolerate this fraud, and I don’t agree with societies and financial firms that make you and I pay for the wrong doing of others. I choose fairness and justice above all.“ (Amir Liberman, 2007)

Is this even allowed?

Of course, there is nothing new about fighting fraud, it is in many places even demanded, and using voice analysis simply makes the process much more friendly and rapid, all within the limitations and restrictions imposed in different territories.

Does InTone.AI comply with the GDPR regulations?

GDPR is a blessed framework designed to protect individual from unjust and abuse of their privacy for no justified reason. It sets several ground principles which all our solutions comply with, such as strong protection of the data collected, encryption and storage, privacy by design… it also sets responsibilities for different stake holders in the analysis process between the data controller and the data processor. The main test in GDPR is the legitimacy and need of the test to comply with the company’s fair and legitimate business needs, and weight those against the private individual right for privacy protection.

It is clear that the insurer has legitimate need to validate the claim and avoid fraud, and with the guards of informed consent and considering voice analysis is the least harassing method to validate the truth, it should pass any GDPR legitimacy test. We can provide some compliance guidelines created for us by top GDPR experts.

What are the benefits for customers?

Befits to the insurance is mainly in the streamlined process of validating claims, right after the first notice of loss the initial warning flag is set either on or off. The insurer can now focus direct resources on the high-risk claims, investigate them from a completely new starting point and indeed, clients reported savings in the ranges of 25-40% of the total expected payout – which has an incredible effect on the combined ratio and the profitability of the company. But when it comes to identifying low risk claims, I think the win is no less important for the insurer and the end client alike. In these majority of the cases, the insurer can repay the client and assist them at their time of need much faster, keep the client happy with great and hassle-free service. Ultimately the improved profitability is returned to the clients with lower premiums, so fighting fraud effectively is a win for everybody.

Which insurance companies do you work with?

While we never discuss clients information I can share we work today with companies in Japan and in the APAC region, Latin America, South Africa and some first activities in Europe.

And in Switzerland?

I’m afraid we don’t operate In Switzerland yet. I really hope this last trip will ignite some new activity as fraud is bad, and it is getting worst.

Wouldn’t this voice analysis be particularly useful for secret services?

Nemesysco is active in many regions in the fields of HLS and anti-terror. It was the main purpose many years ago for Nemesysco’s creation. Emotion Logic is more about to the business and softer side of the activities.

Who is better at detecting lies, a lie detector or your technology?

Professionals in the field of investigations know that there is no such thing as a real “lie-detector” – the only thing all these systems can detect is the arousal or the reaction in the body due to the fear of the lie, but if there is no fear, there is no reaction and then no detection. Excessive fear will also create bad results, and so it is crucial to learn the limits of each system. We consider our solutions ideal for investigations, as they highlight any portion of the conversation that stands out emotionally, allowing the user to really understand what goes on. I think our systems stand out in their ease of use, and remarkable accuracy separating the reaction to the different emotions (which no other tool can do) – as well as picking up on these indications of extreme emotion and “risk” that should be further examined by the SIU.

„Lies are told to change the future. To change the course of things to come…“ See Amir Liberman’s lecture at TEDx Amsterdam in 2012

Please explain to our readers what your second company, Emotion-Logic, does.

Emotion detection and risk assessment are not as trivial as one may think, and at Nemesysco we spend a lot of time training our clients to properly use the solutions and gather insights.
Emotion-Logic was created to fuse modern AI tools to complement our core technology, and provide immediate insights to our users in every use case.

To achieve this goal, we’ve added speech to text, noise cleaners, and various statistical and AI tools to automate and simplify everything that can be simplified, and provide data-rich and most accurate indications. For example, one of the new services we provide via Emotion Logic platform is our FeelGPT advisors. These AI based virtual professionals are made to augment your professional staff with top experts equipped with the emotion detection superpower, and restless attitude to assist the company’s business goals. We have FeelGPTs for fraud detection, for compliance assurance, for sales optimizations and for customer satisfaction – to mention only those in the insurance space.

We also created services such as AppTone that serves our clients by sending whatsapp questionnaires for any purpose, from completing (and risk assessing) claims information post the FNOL call, to assist underwriting processes, loan taking, or even market surveys. In all these services we merge the speech-to-text data with the genuine emotion analysis to create best-in-class automation for insights summarization and risk analysis. The power of AI in really understanding the combined protocol is an exciting new possibility we never had before the recent year. With an ear to the needs of privacy and data protection concerns, we also offer on-prem installations of the Emotion Logic technology ensuring that no data ever leaves the insurers premises.

Who is it aimed at?

Emotion Logic was first created with developers in mind and was created as a set of programming tools. As the development of the platform became more rapid, additional options matured, such as the FeelGPT and the AppTone, opening immediate access to the platform and the technology for users from the HR, insurance and call centers.

Emotion Logic uses voice analysis to recognize emotions for personality and risk assessment. In which areas is this used?

Emotion Logic and its parent company Nemesysco both work with the technology I started developing back in 1997, mainly for security reasons. While Nemesysco is still dominant in the HLS and intelligence, as the time passed, we’ve discovered different domains that could benefit from genuine emotion detection and objective risk assessment, which I think is the most accurate description of what the technology does. These fields vary across so many directions, from Insurance fraud detection to HR recruitment, personality and surveys, call centers quality of service assessments, training and education, medical applications and even TV shows (we were part of the Big Brother shows in Germany, Belgium and other places…).

Tell us more about the Big Brother shows.

I was fortunate enough to come across the relevant talented people at Endemol that saw immediately the potential of the technology in the TV space, and we worked together for several years in Australia (that was the first), followed by Germany, India and several other locations. The technology was used to test the honesty of the house mates in various situations, and due to the special needs of TV making, we created a special simple to use software to show the likelihood of lie telling with 5 bricks of green to red indications.  

Finally, this question: What percentage of your analyses are correct?

While field research conducted in highly controlled environments showed accuracy rates (both true positive and true negative) in ratios well over 90%, the accuracy of emotion detection and risk assessment is highly dependent on the circumstances and test setting, and so it will be misleading to give definitive numbers. The technology and detections are incredibly accurate in real-life settings, when the need to hide to truth is genuine.

Read also: Swiss Insurtech Hub: „2nd Insur Tech Ecosystem Meetup“

Amir Liberman was born in Jerusalem, Israel in 1972. Amir is a leading researcher in the field of human voice analysis, and the first to identify the 3 basic sensations in the human voice: Excitement (positive sensation), Stress (negative sensation) and uncertainty (cognitivestress).

Amir began his voice analysis research in 1997, in response to a terror attack that took the life of 3 young mothers in Israel. His original intention was to build the „ultimate lie detector“ based on speech analysis and other more common techniques of veracity assessment, but quickly changed the scope of his research to explore unique properties of the voice he identified in his preliminary test cases.

Liberman’s first voice related patent was published in 1998, summarizing a research of 2 years, leading the way to the development of the first ever commercial computerized emotion detector. His later discoveries were published in his second voice analysis patent from 1999, identifying “Concentration”, “Anticipation” and “Arousal” (the “Love Detector” patent). Amir was the first to address the different approach needed in assessing the apparent Rapport with a person, in contrast to his true underlying feelings – which cannot be done by the traditional phonetic measurements.

Being a self-educated & self-funded researcher, Liberman’s methods of research are unique and unfamiliar to the world of traditional phonetics, as most of his research is done in real-life settings, not in a laboratory atmosphere.

Amir formed Nemesysco, Ltd. in April of 2000 to manage his IP rights and development projects. Since then, he has invested all his efforts and resources in perfecting and fine-tuning LVA technology and its applications for home-land security needs, fraud prevention, call centers utilities and CRM applications. In his discovery first published in 2005, Amir was the first to identify and classify „Anger“, „Happy“ and „Upset“ in the voice, 3 of the most important emotions in call center settings.

---

---