{"id":7720,"date":"2023-04-09T12:51:00","date_gmt":"2023-04-09T12:51:00","guid":{"rendered":"https:\/\/www.lemonade.com\/blog\/?p=7720"},"modified":"2023-07-25T06:06:02","modified_gmt":"2023-07-25T06:06:02","slug":"ai-can-vanquish-bias","status":"publish","type":"post","link":"https:\/\/www.lemonade.com\/blog\/ai-can-vanquish-bias\/","title":{"rendered":"AI Can Vanquish Bias"},"content":{"rendered":"

Insurance is the business of assessing risks and pricing policies to match. As no two people are entirely alike, that means treating different people\u00a0differently.\u00a0But how to segment people without discriminating unfairly<\/em>?<\/p>\n

Thankfully, no insurer will ever use membership in a ‘protected class’ (race, gender, religion…) as a pricing factor. It\u2019s illegal, unethical, and unprofitable. But while that sounds like the end of the matter, it\u2019s not.<\/p>\n

Take your garden-variety \u2018credit score.\u2019 Credit scores are derived from objective\u00a0data that don\u2019t include race<\/a>, and are highly predictive of insurance losses<\/a>. What\u2019s not to like?\u00a0Indeed, most\u00a0regulators allow the use of credit-based insurance scores, and in the US these can impact your premiums by up to 288%<\/a>. But it turns out there is\u00a0something not to like: credit scores are also highly predictive of skin color, acting in effect as a proxy for race<\/a>. For this reason, California<\/a>, Massachusetts, and Maryland don\u2019t allow insurance pricing based on credit scores.<\/p>\n

Reasonable people may disagree on whether credit scores discriminate fairly or unfairly – and we can have that debate because we can all get our heads around the question at hand. Credit scores are a 3 digit\u00a0number, derived from a static formula that weighs 5 self-explanatory factors<\/a>.<\/p>\n

But in the era of big data and artificial intelligence, all that could change. AI crushes humans\u00a0at chess, for example, because it uses algorithms that no human could create, and none fully understand. The AI encodes its own fabulously intricate instructions, using billions of data to train its machine learning engine. Every time it plays (and it plays millions of times a day), the machine learns, and the algorithm morphs.<\/p>\n

What happens when those capabilities are harnessed for assessing risk and pricing insurance?<\/p>\n

Many fear that such ‘black box’ systems will make matters worse, producing the kind of proxies for race that credit scores do, but without giving us the ability to scrutinize and regulate them. If 5 factors mimic race unwittingly, they say, imagine how much worse it will be in the era of big data!<\/p>\n

But while it\u2019s easy to be alarmist, machine learning and big data are more likely to solve\u00a0the \u2018credit score problem\u2019 than to compound it. You see, problems\u00a0that arise while using 5 factors aren\u2019t multiplied<\/em>\u00a0by millions of data – they are divided<\/em>\u00a0by them.<\/p>\n

To understand why, let\u2019s think about the process of using data to segment – or \u2018discriminate\u2019 – as evolving in 3 phases.<\/p>\n

\"\"<\/p>\n

Phase 1:<\/h3>\n

In Phase 1 all\u00a0people are treated as though they are identical. Everyone represents the same risk, and is therefore charged the same premium (per unit of coverage). This was commonplace in insurance until the 18th century.<\/p>\n

\"\"<\/p>\n

Phase 1 avoids discriminating based on race, ethnicity, gender, religion or anything else for that matter, but that doesn\u2019t make it fair, practical, or even legal.<\/p>\n

One problem with Phase 1 is that people who are more thoughtful and careful are made to subsidize those who are more thoughtless and careless. Externalizing the costs of risky behavior doesn\u2019t make for good policy, and isn\u2019t fair to those who are stuck with the bill.<\/p>\n

Another problem is that it\u2019s no longer workable, because people who are better-than-average risks will seek lower prices elsewhere – leaving the insurer with average premiums, but riskier-than-average customers (a problem known as \u2018adverse selection\u2019).<\/p>\n

Finally, best intentions notwithstanding, Phase 1 fits the legal textbook definition of ‘unfair discrimination.’ The law<\/a> mandates that, subject to “practical limitations,” a price is “unfairly discriminatory\u201d if it \u201cfails to reflect with reasonable accuracy the differences in expected losses.\u201d In other words, within the confines of what\u2019s \u2018practical,\u2019 insurers must<\/em>\u00a0charge each person a rate that\u2019s proportionate to their risk.<\/em><\/p>\n

Which brings us to Phase 2.<\/p>\n

Phase 2:<\/h3>\n

Phase 2 sees the population divided into subgroups according to their risk profile. This process is data-driven and impartial, yet as the data are relatively basic, the groupings are relatively crude. Phase 2 – broadly speaking – reflects the state of the industry today, and it\u2019s far from ideal.<\/p>\n

\"\"<\/p>\n

Sorting with limited data generates relatively\u00a0few, large groups, and two big problems.<\/p>\n

The first is that the groups may proxy protected classes. Take gender as an example. Imagine, if you will, that women are – on average – better risks than men (say the average\u00a0risk score for a woman is 40,\u00a0on a 1-100 scale, and 60 for men). We\u2019d still expect many<\/em>\u00a0women to be sub-average risks, and many<\/em>\u00a0men to be better than average.<\/p>\n

So while crude groupings may be statistically sound, Phase 2 might penalize low-risk men by tarring all<\/em>\u00a0men with the same brush.<\/p>\n

The second problem is that – even if the groups don\u2019t represent protected classes – responsible members of the group are still made to pay more (per unit of risk) than their less responsible compatriots. That\u2019s what happens when you impose a uniform rate on a nonuniform group. As we saw, this is the textbook definition of \u2018unfair discrimination,\u2019 which we tolerate as a necessary evil, born of \u2018practical limitations.\u2019 But the practical limitations of yesteryear are crumbling, and there\u2019s a four letter word for a \u2018necessary evil\u2019 that is no longer necessary\u2026<\/p>\n

Which brings us to Phase 3.<\/p>\n

Phase 3:<\/h3>\n

Phase 3 continues where Phase 2 ends: breaking monolithic groups into component subgroups.<\/p>\n

\"\"<\/p>\n

It does it on a massive scale, using orders of magnitude more data, which machine learning crunches to produce very complex multivariate risk scores. The upshot is that today\u2019s coarse groupings are relentlessly shrunk, until – ultimately – each person is a \u2018group of one.\u2019 A grouping that in Phase 2 might be a proxy for \u2018men,\u2019 and scored as a \u201860,\u2019 is now seen as a series of individuals, some with a risk score of 90, others of 30, and so forth. This series still averages<\/em>\u00a0a score of 60 – but while that average may be applied to all men in Phase 2, it\u2019s applied to none<\/em>\u00a0of them in Phase 3.<\/p>\n

In Phase 3, large groups crumble under the weight of the data and the crushing power of the machine.\u00a0Insurance remains the business of pooling premiums to pay claims, but now each person contributes to the pool in direct proportion to the risk they<\/em>\u00a0represent – rather than the risk represented by a large group of somewhat similar people. By charging every person the same, per unit of risk<\/em>, we sidestep the inequity, illegality, and the moral hazard of charging the careful to pay for the careless, and of grouping people in ways that proxy race, gender, or religion. It\u2019s like we said: problems\u00a0that arise while using 5 factors aren\u2019t multiplied<\/em>\u00a0by millions of data – they are divided<\/em>\u00a0by them.<\/p>\n

\"\"<\/p>\n

Insurance Can Tame AI<\/h3>\n

It\u2019s encouraging to know that Phase 3 has the potential to make insurance more fair, but how can we audit the algorithm to ensure it actually lives up to this promise? There’s been some progress towards ‘explainability’ in Machine Learning, but without true transparency into that black box, how are we to assess the impartiality of its outputs?<\/p>\n

By their outcomes.<\/p>\n

But we must tread gingerly, and check our intuitions at the door. It\u2019s tempting to say that an algorithm that charges women more than men, or black people more than white people, or Jews more than gentiles – is discriminating unfairly. That\u2019s the obvious conclusion, the traditional one, and – in Phase 3 – it\u2019s likely to be the wrong one.<\/p>\n

Let\u2019s say I am Jewish (I am), and that part of my tradition involves lighting a bunch of candles throughout the year (it does). In our home we light candles every Friday night, every holiday eve, and we\u2019ll burn through about two hundred candles over the 8 nights of Hanukkah. It would not be surprising if I, and others like me, represented a higher risk of fire than the national average. So, if the AI charges Jews, on average<\/em>, more than non-Jews for fire insurance, is that unfairly discriminatory?<\/p>\n

It depends.<\/p>\n

It would definitely be a problem if being Jewish, per se, resulted in higher premiums whether or not you\u2019re the candle-lighting kind of Jew<\/em>. Not all Jews are avid candle lighters, and an algorithm that treats all Jews like the \u2018average Jew,\u2019 would be despicable. That, though, is a Phase 2 problem.<\/p>\n

A Phase 3 algorithm that identifies people\u2019s proclivity for candle lighting, and charges them more for the risk that this penchant actually represents<\/em>, is entirely fair. The fact that such a fondness for candles is unevenly distributed in the population, and more highly concentrated among Jews, means that, on average<\/em>,\u00a0Jews will pay more. It does not\u00a0mean that people are charged more for being Jewish.<\/em><\/p>\n

It\u2019s hard to overstate the importance of this distinction. All cows have four legs, but not all things with four legs are cows.<\/p>\n

The upshot is that the mere fact that an algorithm charges Jews – or women, or black people – more on average\u00a0does not render it unfairly discriminatory. Phase 3 doesn\u2019t do averages. In common with Dr. Martin Luther King, we dream of living in a world where we are judged by\u00a0the content of our character. We want to be assessed as individuals<\/em>, not by reference to our racial, gender, or religious markers. If the AI is treating us all this way, as humans, then it is being fair. If I\u2019m charged more for my candle-lighting habit, that\u2019s as it should be, even if the behavior I\u2019m being charged for is disproportionately common among Jews. The AI is responding to my fondness for candles (which is a real risk factor), not to my tribal affiliation (which is not).<\/p>\n

So if differential<\/em>\u00a0pricing isn\u2019t proof of unfair pricing, what is? What \u2018outcome\u2019 is the telltale sign of unfair discrimination in Phase 3?<\/p>\n

Differential loss ratios.<\/p>\n

The \u2018pure loss ratio\u2019 is the ratio of the dollars paid out in claims by the insurance company, to the dollars it collects in premiums. If an insurance company charges all customers a rate proportionate to the risk they pose, this ratio should be constant across their customer base. We\u2019d expect to see fluctuations among individuals, sure, but once we aggregate people into sizable groupings – say by gender, ethnicity or religion – the law of large numbers should kick in, and we should see a consistent loss ratio across such cohorts. If that\u2019s the case, that would suggest that even if certain groups – on average – are paying more, these higher rates are fair, because they represent commensurately higher claim payouts. A system is fair – by law<\/em>\u00a0– if each of us is paying in direct proportion to the risk we represent. This is what the proposed \u2018Uniform Loss Ratio\u2019 (ULR)\u00a0test, tests. It puts insurance in the enviable position of being able to keep AI honest with a simple, objective, and easily administered test.<\/p>\n

It is possible, of course, for an insurance company to charge a fair premium, but then have a bias when it comes to paying claims. The beauty of the ULR test is that such a bias would be readily exposed. Simply put, if certain groups have a lower loss ratio than the population at large, that would signal that they are being treated unfairly. Their rates are too high, relative to the payout they are receiving.<\/p>\n

\"\"<\/p>\n

ULR helps us overcome another major concern with AI. Even though machines do not have inherent<\/em> biases, they can inherit<\/em>\u00a0biases. Imagine, for example, that the machine finds that people who are arrested are also more likely to be robbed. I have no idea whether this is the case, but it wouldn\u2019t be a shocking discovery. Prior run-ins with the police would, in this hypothetical, become a legitimate factor in assessing property-insurance premiums. So far, so objective.<\/p>\n

The problem arises if some of the arresting officers are themselves biased, leading – for example – to an elevated rate of black people being arrested for no good reason. If that were the case, the rating algorithm would inherit the humans\u2019 racial bias: a person wouldn\u2019t pay more insurance premiums for being black, per se<\/em>, but they would pay more for being arrested – and the likelihood of that<\/em>\u00a0happening would be heightened for black people.<\/p>\n

While my example is hypothetical, the problem is very real. Worried about AI-inherited biases, many people are understandably sounding the retreat. The better response, though, is to sound the advance.<\/p>\n

You see, machines can overcome the biases that contaminate their training data if they can continuously calibrate their algorithms against unbiased\u00a0<\/em>data. In insurance, ULR provides such a true north. Applying the ULR test, the AI would quickly determine that having \u2018been arrested\u2019 isn\u2019t equally predictive of claims across the population. As data accumulate, the \u2018been arrested\u2019 group would subdivide, because the AI would detect that for certain people being arrested is less predictive of future claims than it is for others. The algorithm would self-correct, adjusting the weighting of this datum to compensate for human bias.<\/p>\n

(When a system is accused of bias, the go-to defense runs something like: \u201cbut we don\u2019t even collect gender, race, religion or sexual preference.\u201d Such indignation is doubly misplaced. For one, as we\u2019ve seen, systems can be prejudiced without direct knowledge of these factors. For another, the best way for ULR-calibrated-systems to neutralize bias, is to actually know\u00a0these factors.)<\/p>\n

Bottom line: problems that arise while using 5 factors aren\u2019t multiplied<\/em>\u00a0by millions of data – they are divided<\/em>\u00a0by them.<\/p>\n

The Machines Are Coming. Look Busy.<\/h3>\n

Phase 3 doesn\u2019t exist yet, but it\u2019s a future we should embrace and prepare for. That\u00a0requires insurance companies to re-architect their customer journey to be entirely digital, and reconstitute their systems and processes on an AI substrate. In many jurisdictions, it also requires a rethinking of the way insurance pricing is regulated. Adopting the ULR test would be a big step forward. In Europe, the regulatory framework could become Phase-3-ready with minor tweaks. In the US, the process of filing rates in a simple and static multiplication chart for human review doesn\u2019t scale as we transition from Phase 2 to 3. At a minimum, regulators should allow these lookup-tables to include a column for a black box \u2018risk factor.\u2019 The ULR test would ensure these never cause more harm than good, while this additional pricing factor would enable emerging technologies to benefit insurers and insureds alike.<\/p>\n

Nice to meet you<\/h3>\n

When we meet someone for the first time, we tend to lump them with others with whom they share surface similarities. It\u2019s human nature, and it can be unfair. Once we learn more about that individual, superficial judgments hopefully give way to a merits-based assessment. It\u2019s a welcome progression, and it’s powered by intelligence and data.<\/p>\n

What intelligence and data have done for humanity throughout our history, artificial intelligence and big data can start to do for the insurance industry. This is not only increasingly possible as a matter of technology, it is also desirable<\/em>\u00a0as a matter of policy. Furthermore, as it will represent a huge competitive advantage, it is also largely inevitable. Those who fail to embrace the precision underwriting and pricing of Phase 3 will ultimately be adversely-selected out of business.<\/p>\n

Insurance is the business of assessing risks, and pricing policies to match. As no two people are entirely alike, that means treating different people\u00a0differently. For the first time in history, we\u2019re on the cusp of being able to do precisely<\/em>\u00a0that.<\/p>\n","protected":false},"excerpt":{"rendered":"

Algorithms we may never understand can be trusted to price insurance in a way that is far more precise, and far more fair, than today\u2019s human equivalents.<\/p>\n","protected":false},"author":11,"featured_media":7721,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[10,12],"tags":[],"puppies_section":[],"class_list":{"0":"post-7720","1":"post","2":"type-post","3":"status-publish","4":"format-standard","5":"has-post-thumbnail","7":"category-news","8":"category-transparency","9":"post-hentry"},"acf":[],"yoast_head":"\nAI Can Vanquish Bias - Lemonade Blog<\/title>\n<meta name=\"description\" content=\"Algorithms we may never understand can be trusted to price insurance in a way that is far more precise - and far more fair - than today\u2019s human equivalents.\" \/>\n<meta name=\"robots\" content=\"index, follow, max-snippet:-1, max-image-preview:large, max-video-preview:-1\" \/>\n<link rel=\"canonical\" href=\"https:\/\/www.lemonade.com\/blog\/ai-can-vanquish-bias\/\" \/>\n<meta property=\"og:locale\" content=\"en_US\" \/>\n<meta property=\"og:type\" content=\"article\" \/>\n<meta property=\"og:title\" content=\"AI Can Vanquish Bias\" \/>\n<meta property=\"og:description\" content=\"Algorithms we may never understand can be trusted to price insurance in a way that is far more precise - and far more fair - than today\u2019s human equivalents.\" \/>\n<meta property=\"og:url\" content=\"https:\/\/www.lemonade.com\/blog\/ai-can-vanquish-bias\/\" \/>\n<meta property=\"og:site_name\" content=\"Lemonade Blog\" \/>\n<meta property=\"article:publisher\" content=\"https:\/\/www.facebook.com\/Lemonade\/\" \/>\n<meta property=\"og:image\" content=\"https:\/\/www.lemonade.com\/blog\/wp-content\/uploads\/2019\/12\/OG-1200x628px-2.jpg\" \/>\n\t<meta property=\"og:image:width\" content=\"1200\" \/>\n\t<meta property=\"og:image:height\" content=\"628\" \/>\n\t<meta property=\"og:image:type\" content=\"image\/jpeg\" \/>\n<meta name=\"author\" content=\"Daniel Schreiber\" \/>\n<meta name=\"twitter:card\" content=\"summary_large_image\" \/>\n<meta name=\"twitter:title\" content=\"AI Can Vanquish Bias\" \/>\n<meta name=\"twitter:description\" content=\"Algorithms we may never understand can be trusted to price insurance in a way that is far more precise - and far more fair - than today\u2019s human equivalents.\" \/>\n<meta name=\"twitter:image\" content=\"https:\/\/www.lemonade.com\/blog\/wp-content\/uploads\/2019\/12\/OG-1200x628px-2.jpg\" \/>\n<meta name=\"twitter:creator\" content=\"@daschreiber\" \/>\n<meta name=\"twitter:site\" content=\"@lemonade_inc\" \/>\n<meta name=\"twitter:label1\" content=\"Written by\" \/>\n\t<meta name=\"twitter:data1\" content=\"Daniel Schreiber\" \/>\n\t<meta name=\"twitter:label2\" content=\"Est. reading time\" \/>\n\t<meta name=\"twitter:data2\" content=\"13 minutes\" \/>\n<script type=\"application\/ld+json\" class=\"yoast-schema-graph\">{\"@context\":\"https:\\\/\\\/schema.org\",\"@graph\":[{\"@type\":\"Article\",\"@id\":\"https:\\\/\\\/www.lemonade.com\\\/blog\\\/ai-can-vanquish-bias\\\/#article\",\"isPartOf\":{\"@id\":\"https:\\\/\\\/www.lemonade.com\\\/blog\\\/ai-can-vanquish-bias\\\/\"},\"author\":{\"name\":\"Daniel Schreiber\",\"@id\":\"https:\\\/\\\/www.lemonade.com\\\/blog\\\/#\\\/schema\\\/person\\\/124a6b9b02ded1236b63a3dabb69293a\"},\"headline\":\"AI Can Vanquish Bias\",\"datePublished\":\"2023-04-09T12:51:00+00:00\",\"dateModified\":\"2023-07-25T06:06:02+00:00\",\"mainEntityOfPage\":{\"@id\":\"https:\\\/\\\/www.lemonade.com\\\/blog\\\/ai-can-vanquish-bias\\\/\"},\"wordCount\":2616,\"commentCount\":0,\"image\":{\"@id\":\"https:\\\/\\\/www.lemonade.com\\\/blog\\\/ai-can-vanquish-bias\\\/#primaryimage\"},\"thumbnailUrl\":\"https:\\\/\\\/www.lemonade.com\\\/blog\\\/wp-content\\\/uploads\\\/2019\\\/12\\\/Blog-Featured-image-1553x728.jpg\",\"articleSection\":[\"news\",\"transparency\"],\"inLanguage\":\"en-US\",\"potentialAction\":[{\"@type\":\"CommentAction\",\"name\":\"Comment\",\"target\":[\"https:\\\/\\\/www.lemonade.com\\\/blog\\\/ai-can-vanquish-bias\\\/#respond\"]}]},{\"@type\":\"WebPage\",\"@id\":\"https:\\\/\\\/www.lemonade.com\\\/blog\\\/ai-can-vanquish-bias\\\/\",\"url\":\"https:\\\/\\\/www.lemonade.com\\\/blog\\\/ai-can-vanquish-bias\\\/\",\"name\":\"AI Can Vanquish Bias - Lemonade Blog\",\"isPartOf\":{\"@id\":\"https:\\\/\\\/www.lemonade.com\\\/blog\\\/#website\"},\"primaryImageOfPage\":{\"@id\":\"https:\\\/\\\/www.lemonade.com\\\/blog\\\/ai-can-vanquish-bias\\\/#primaryimage\"},\"image\":{\"@id\":\"https:\\\/\\\/www.lemonade.com\\\/blog\\\/ai-can-vanquish-bias\\\/#primaryimage\"},\"thumbnailUrl\":\"https:\\\/\\\/www.lemonade.com\\\/blog\\\/wp-content\\\/uploads\\\/2019\\\/12\\\/Blog-Featured-image-1553x728.jpg\",\"author\":{\"@id\":\"https:\\\/\\\/www.lemonade.com\\\/blog\\\/#\\\/schema\\\/person\\\/124a6b9b02ded1236b63a3dabb69293a\"},\"description\":\"Algorithms we may never understand can be trusted to price insurance in a way that is far more precise - and far more fair - than today\u2019s human equivalents.\",\"breadcrumb\":{\"@id\":\"https:\\\/\\\/www.lemonade.com\\\/blog\\\/ai-can-vanquish-bias\\\/#breadcrumb\"},\"inLanguage\":\"en-US\",\"potentialAction\":[{\"@type\":\"ReadAction\",\"target\":[\"https:\\\/\\\/www.lemonade.com\\\/blog\\\/ai-can-vanquish-bias\\\/\"]}]},{\"@type\":\"ImageObject\",\"inLanguage\":\"en-US\",\"@id\":\"https:\\\/\\\/www.lemonade.com\\\/blog\\\/ai-can-vanquish-bias\\\/#primaryimage\",\"url\":\"https:\\\/\\\/www.lemonade.com\\\/blog\\\/wp-content\\\/uploads\\\/2019\\\/12\\\/Blog-Featured-image-1553x728.jpg\",\"contentUrl\":\"https:\\\/\\\/www.lemonade.com\\\/blog\\\/wp-content\\\/uploads\\\/2019\\\/12\\\/Blog-Featured-image-1553x728.jpg\",\"width\":1553,\"height\":728},{\"@type\":\"BreadcrumbList\",\"@id\":\"https:\\\/\\\/www.lemonade.com\\\/blog\\\/ai-can-vanquish-bias\\\/#breadcrumb\",\"itemListElement\":[{\"@type\":\"ListItem\",\"position\":1,\"name\":\"Home\",\"item\":\"https:\\\/\\\/www.lemonade.com\\\/blog\\\/\"},{\"@type\":\"ListItem\",\"position\":2,\"name\":\"AI Can Vanquish Bias\"}]},{\"@type\":\"WebSite\",\"@id\":\"https:\\\/\\\/www.lemonade.com\\\/blog\\\/#website\",\"url\":\"https:\\\/\\\/www.lemonade.com\\\/blog\\\/\",\"name\":\"Lemonade Blog\",\"description\":\"Talking innovation, purpose, user experience and insurance\",\"potentialAction\":[{\"@type\":\"SearchAction\",\"target\":{\"@type\":\"EntryPoint\",\"urlTemplate\":\"https:\\\/\\\/www.lemonade.com\\\/blog\\\/?s={search_term_string}\"},\"query-input\":{\"@type\":\"PropertyValueSpecification\",\"valueRequired\":true,\"valueName\":\"search_term_string\"}}],\"inLanguage\":\"en-US\"},{\"@type\":\"Person\",\"@id\":\"https:\\\/\\\/www.lemonade.com\\\/blog\\\/#\\\/schema\\\/person\\\/124a6b9b02ded1236b63a3dabb69293a\",\"name\":\"Daniel Schreiber\",\"image\":{\"@type\":\"ImageObject\",\"inLanguage\":\"en-US\",\"@id\":\"https:\\\/\\\/secure.gravatar.com\\\/avatar\\\/77614b4021e3c63f82d4b62a2407a4fff4111f53235046d5e55f9324b4705257?s=96&d=mm&r=g\",\"url\":\"https:\\\/\\\/secure.gravatar.com\\\/avatar\\\/77614b4021e3c63f82d4b62a2407a4fff4111f53235046d5e55f9324b4705257?s=96&d=mm&r=g\",\"contentUrl\":\"https:\\\/\\\/secure.gravatar.com\\\/avatar\\\/77614b4021e3c63f82d4b62a2407a4fff4111f53235046d5e55f9324b4705257?s=96&d=mm&r=g\",\"caption\":\"Daniel Schreiber\"},\"sameAs\":[\"https:\\\/\\\/x.com\\\/daschreiber\"]}]}<\/script>\n<!-- \/ Yoast SEO plugin. -->","yoast_head_json":{"title":"AI Can Vanquish Bias - Lemonade Blog","description":"Algorithms we may never understand can be trusted to price insurance in a way that is far more precise - and far more fair - than today\u2019s human equivalents.","robots":{"index":"index","follow":"follow","max-snippet":"max-snippet:-1","max-image-preview":"max-image-preview:large","max-video-preview":"max-video-preview:-1"},"canonical":"https:\/\/www.lemonade.com\/blog\/ai-can-vanquish-bias\/","og_locale":"en_US","og_type":"article","og_title":"AI Can Vanquish Bias","og_description":"Algorithms we may never understand can be trusted to price insurance in a way that is far more precise - and far more fair - than today\u2019s human equivalents.","og_url":"https:\/\/www.lemonade.com\/blog\/ai-can-vanquish-bias\/","og_site_name":"Lemonade Blog","article_publisher":"https:\/\/www.facebook.com\/Lemonade\/","og_image":[{"width":1200,"height":628,"url":"https:\/\/www.lemonade.com\/blog\/wp-content\/uploads\/2019\/12\/OG-1200x628px-2.jpg","type":"image\/jpeg"}],"author":"Daniel Schreiber","twitter_card":"summary_large_image","twitter_title":"AI Can Vanquish Bias","twitter_description":"Algorithms we may never understand can be trusted to price insurance in a way that is far more precise - and far more fair - than today\u2019s human equivalents.","twitter_image":"https:\/\/www.lemonade.com\/blog\/wp-content\/uploads\/2019\/12\/OG-1200x628px-2.jpg","twitter_creator":"@daschreiber","twitter_site":"@lemonade_inc","twitter_misc":{"Written by":"Daniel Schreiber","Est. reading time":"13 minutes"},"schema":{"@context":"https:\/\/schema.org","@graph":[{"@type":"Article","@id":"https:\/\/www.lemonade.com\/blog\/ai-can-vanquish-bias\/#article","isPartOf":{"@id":"https:\/\/www.lemonade.com\/blog\/ai-can-vanquish-bias\/"},"author":{"name":"Daniel Schreiber","@id":"https:\/\/www.lemonade.com\/blog\/#\/schema\/person\/124a6b9b02ded1236b63a3dabb69293a"},"headline":"AI Can Vanquish Bias","datePublished":"2023-04-09T12:51:00+00:00","dateModified":"2023-07-25T06:06:02+00:00","mainEntityOfPage":{"@id":"https:\/\/www.lemonade.com\/blog\/ai-can-vanquish-bias\/"},"wordCount":2616,"commentCount":0,"image":{"@id":"https:\/\/www.lemonade.com\/blog\/ai-can-vanquish-bias\/#primaryimage"},"thumbnailUrl":"https:\/\/www.lemonade.com\/blog\/wp-content\/uploads\/2019\/12\/Blog-Featured-image-1553x728.jpg","articleSection":["news","transparency"],"inLanguage":"en-US","potentialAction":[{"@type":"CommentAction","name":"Comment","target":["https:\/\/www.lemonade.com\/blog\/ai-can-vanquish-bias\/#respond"]}]},{"@type":"WebPage","@id":"https:\/\/www.lemonade.com\/blog\/ai-can-vanquish-bias\/","url":"https:\/\/www.lemonade.com\/blog\/ai-can-vanquish-bias\/","name":"AI Can Vanquish Bias - Lemonade Blog","isPartOf":{"@id":"https:\/\/www.lemonade.com\/blog\/#website"},"primaryImageOfPage":{"@id":"https:\/\/www.lemonade.com\/blog\/ai-can-vanquish-bias\/#primaryimage"},"image":{"@id":"https:\/\/www.lemonade.com\/blog\/ai-can-vanquish-bias\/#primaryimage"},"thumbnailUrl":"https:\/\/www.lemonade.com\/blog\/wp-content\/uploads\/2019\/12\/Blog-Featured-image-1553x728.jpg","author":{"@id":"https:\/\/www.lemonade.com\/blog\/#\/schema\/person\/124a6b9b02ded1236b63a3dabb69293a"},"description":"Algorithms we may never understand can be trusted to price insurance in a way that is far more precise - and far more fair - than today\u2019s human equivalents.","breadcrumb":{"@id":"https:\/\/www.lemonade.com\/blog\/ai-can-vanquish-bias\/#breadcrumb"},"inLanguage":"en-US","potentialAction":[{"@type":"ReadAction","target":["https:\/\/www.lemonade.com\/blog\/ai-can-vanquish-bias\/"]}]},{"@type":"ImageObject","inLanguage":"en-US","@id":"https:\/\/www.lemonade.com\/blog\/ai-can-vanquish-bias\/#primaryimage","url":"https:\/\/www.lemonade.com\/blog\/wp-content\/uploads\/2019\/12\/Blog-Featured-image-1553x728.jpg","contentUrl":"https:\/\/www.lemonade.com\/blog\/wp-content\/uploads\/2019\/12\/Blog-Featured-image-1553x728.jpg","width":1553,"height":728},{"@type":"BreadcrumbList","@id":"https:\/\/www.lemonade.com\/blog\/ai-can-vanquish-bias\/#breadcrumb","itemListElement":[{"@type":"ListItem","position":1,"name":"Home","item":"https:\/\/www.lemonade.com\/blog\/"},{"@type":"ListItem","position":2,"name":"AI Can Vanquish Bias"}]},{"@type":"WebSite","@id":"https:\/\/www.lemonade.com\/blog\/#website","url":"https:\/\/www.lemonade.com\/blog\/","name":"Lemonade Blog","description":"Talking innovation, purpose, user experience and insurance","potentialAction":[{"@type":"SearchAction","target":{"@type":"EntryPoint","urlTemplate":"https:\/\/www.lemonade.com\/blog\/?s={search_term_string}"},"query-input":{"@type":"PropertyValueSpecification","valueRequired":true,"valueName":"search_term_string"}}],"inLanguage":"en-US"},{"@type":"Person","@id":"https:\/\/www.lemonade.com\/blog\/#\/schema\/person\/124a6b9b02ded1236b63a3dabb69293a","name":"Daniel Schreiber","image":{"@type":"ImageObject","inLanguage":"en-US","@id":"https:\/\/secure.gravatar.com\/avatar\/77614b4021e3c63f82d4b62a2407a4fff4111f53235046d5e55f9324b4705257?s=96&d=mm&r=g","url":"https:\/\/secure.gravatar.com\/avatar\/77614b4021e3c63f82d4b62a2407a4fff4111f53235046d5e55f9324b4705257?s=96&d=mm&r=g","contentUrl":"https:\/\/secure.gravatar.com\/avatar\/77614b4021e3c63f82d4b62a2407a4fff4111f53235046d5e55f9324b4705257?s=96&d=mm&r=g","caption":"Daniel Schreiber"},"sameAs":["https:\/\/x.com\/daschreiber"]}]}},"_links":{"self":[{"href":"https:\/\/www.lemonade.com\/blog\/wp-json\/wp\/v2\/posts\/7720","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/www.lemonade.com\/blog\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/www.lemonade.com\/blog\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/www.lemonade.com\/blog\/wp-json\/wp\/v2\/users\/11"}],"replies":[{"embeddable":true,"href":"https:\/\/www.lemonade.com\/blog\/wp-json\/wp\/v2\/comments?post=7720"}],"version-history":[{"count":2,"href":"https:\/\/www.lemonade.com\/blog\/wp-json\/wp\/v2\/posts\/7720\/revisions"}],"predecessor-version":[{"id":22978,"href":"https:\/\/www.lemonade.com\/blog\/wp-json\/wp\/v2\/posts\/7720\/revisions\/22978"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/www.lemonade.com\/blog\/wp-json\/wp\/v2\/media\/7721"}],"wp:attachment":[{"href":"https:\/\/www.lemonade.com\/blog\/wp-json\/wp\/v2\/media?parent=7720"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.lemonade.com\/blog\/wp-json\/wp\/v2\/categories?post=7720"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.lemonade.com\/blog\/wp-json\/wp\/v2\/tags?post=7720"},{"taxonomy":"puppies_section","embeddable":true,"href":"https:\/\/www.lemonade.com\/blog\/wp-json\/wp\/v2\/puppies_section?post=7720"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}