Andrew Lockley – Exponential Investor (United Kingdom) –
Yesterday, we heard about how artificial intelligence (AI) is becoming increasingly important in criminal justice. Today, we’ll look at how similar techniques are applied in a closely-related field: insurance. You may find it creepy, but insurers are keen to digitally-scrutinise your personality. They then use this data to work out how risky you are.
Did you hear about the recent spat between Facebook and Admiral? The insurer wanted to use the vast mountain of data on Facebook to identify people with traits suggesting they’re a lower insurance risk. For example, if you arrange to meet a friend “tomorrow afternoon”, that indicates your mind works in a different way to someone who sends an invite for “1615h tomorrow”. The more precise person is apparently a better insurance risk.
Facebook was having none of it. It was perhaps concerned by the risk that people would “game” their Facebook profile, or perhaps even stay off the site altogether. In this individual case, the risks were probably low – but it’s an example of a “slippery slope” argument. Once one insurer is surreptitiously analysing your Facebook activity, it’s likely that others will follow suit. If all insurers are doing it, it may become difficult for drivers to obtain competitive quotes without giving social media access. That’s not a way of doing business that Facebook was keen to facilitate – and so it clamped down.
This is an important story for your investments. Facebook is a large, publicly-traded company. Even if you don’t hold the stock directly, you may be exposed to it in other ways – such as via your pension. Regardless of how you’re exposed, you should be keeping a careful eye on the opportunities and risks posed by these large tech stocks. Fortunately, we’re able to help you with that – and Eoin Treacy’s Frontier Tech Investor is the publication you need to be reading. You can find out more about it here.
Personality analysis is a fascinating space, and one I’ve worked actively in. The market extends across a huge range of online activity: insurance, loans, and recruitment. Even ecommerce can benefit – as the brands you choose are heavily influenced by your psychological traits.
We don’t like to treat issues superficially at Exponential Investor – so today we’re doing a deep dive, with Benoit Allibe from ZenWeShare.
AL: Hi Benoit. Can you start off by telling me a bit about ZenWeShare?
BA: ZenWeShare is a French company that focuses on adding value with personal data. Two years ago, we developed a portable reputation profile for individuals. We noticed that many sharing-economy sites had reputation scores – brands like Airbnb, BlaBlaCar, Ebay, etc. The problem was that a great score on one site didn’t carry across to others. Our product was designed to help people show their existing sharing-economy reputation on every website. This helps users to sell more quickly, and for a better price. Developing this product made us realise that personality traits can be used far beyond the obvious sharing economy sites. For instance, someone who’s regarded as a safe driver by BlaBlaCar users is also likely to be a good insurance risk. This led us into building a technology solution that can assess personality traits, based on existing online personal data. These are of course accessed with user permission.
AL: What kind of personal data are accessed by ZenWeShare?
BA: We only use data that users share with us. It can be the content of their Facebook profiles, as well as sharing economy reputations, or even browsing data. Every bit of data is interesting, when you’re trying to build a better picture of someone. This process can be really rewarding for users, if you can find and build the right use case for them. For instance, it could save hundreds of pounds on car insurance premiums for young drivers, or help people get the job of their dreams. For businesses, it helps go beyond the “one-size-fits-all” approach in marketing and communications – emphasising different benefits to different users.
AL: So you are creating a new way of describing a user online?
BA: Exactly. We provide a way to summarise information about someone, based on their online behaviour. This information can then be used to offer special discounts, adapt insurance premiums, tailor communication with people, and ultimately personalise and improve the interaction between people and businesses.
AL: You mentioned car insurance. You tell me the price can depend on your personality traits. Can it lead to a sort of discrimination?
BA: If you call price differences discrimination, the answer is yes. In fact, every field of a car insurance subion form leads to price differences. That is because statisticians have seen some differences in crash probability, depending on some information about people. They try to reflect these differences in price to propose lower prices to less “statistically risky” people. Think about age, work or postcode. The question is – what is acceptable as a basis for a price change? We have seen that EU has forbidden insurers to base their price on gender, even if insurers know that women have less serious crashes, and therefore deserve lower premiums. The question about using personality traits for adapting premiums is still open, because it is new. Only society has the ultimate answer.
The correlation between personality traits and car accidents has been proven by numerous studies. However, its social and legal acceptability as an insurance price variable is a new question – and one that has no clear answer yet.
A few weeks ago, Admiral tried to roll out an insurance product, based on personality scoring. Unfortunately, its acceptance by society has not been measured – because Facebook forbade the insurer from launching as planned.
In cases outside insurance, the use of personality measurements is widely accepted. For instance – it’s widely used in recruitment, where asking people to fill a questionnaire about their personality is considered acceptable, even routine.
AL: How does your service help the companies that use it?
BA: Based on the information we provide, corporations can adapt their service to their customers. That includes things such as communication tailoring, or price changes. Banks and insurance companies have been changing prices for a very long time. They adapt the interest rate or premium depending on information they have on their customers, such as income, occupation, and age. This is because they have historically seen that some types of people are less risky than others, and they therefore deserve lower prices. This kind of price adaptation is generally accepted, but it can be considered as discrimination in some cases. For instance, in the EU, gender-based price discrimination is unlawful. We’re basically adding a more scientific approach to an old process.
AL: Personality traits seem to be very interesting for many industries. Do you think that it will become publicly available information in the future, such as credit scores are today?
BA: We don’t think that personal data should be made public without user consent, and the same applies to processed data like personality scores. Scoring has to be controllable, transparent and specific to be acceptable.
An extreme view of a bad system of scoring can be seen in China, where they have plans to score every citizen. In these cases, the scoring is centralised by a political authority. It’s not under user consent, and seems to be public to everyone. In fact, the plot of Black Mirror reflects the real situation in China. In a one-party state, that’s quite a worrying situation for freedoms and human rights.
From our side, and from the view of many startups in our field, users have to have control on the information they share. Also, unlike China’s citizen scoring, our results are specific to particular activities. We’re not trying to quantify differences between people, not determine who’s the best citizen.
Please do send your thoughts to firstname.lastname@example.org – we’ll be carefully scoring your responses, to work out if you’re the kind of person we want as a reader.