Insuretech and Beyond – An Evolving Litigation Landscape
In recent years, the life insurance industry has greatly enhanced the speed and efficiency of its underwriting decisions. This change in the underwriting process is due in large part to the use of new technology in data collection and risk assessment. As the utilization of such technology expands, the changes within the industry will accelerate at an even faster rate— potentially before the industry is able to determine whether their changes are ultimately beneficial. Accordingly, the insurance industry should analyze the potential benefits and drawbacks on new technology used to increase underwriting efficiency while adaptive measures are still possible. This analysis should consider both the immediate practical concerns of implementing this technology and the larger impact such technology will have on the industry at large.
This paper highlights the recent technological changes within the life insurance landscape, and the legal issues implicated by its utilization, including rescission. It is not intended to be an exhaustive discussion of rescission or any legal issue but rather, illustrative of those implicated by the technology changes and trends. The issues are being litigated now and the litigation will undoubtedly escalate.
Move Towards Accelerated Underwriting
The underwriting of life insurance policies requires insurers to gather extensive information on applicants to determine their individual mortality risks.1 Such a nebulous concept requires the analysis of personalized information, including an individual applicant’s medical background, family history, driving records, and general lifestyle activities.2 The accumulation of this information has traditionally been a lengthy process, requiring applicants to undergo several medical evaluations and to procure and submit a significant amount of documentation.3 This underwriting process often led to general customer dissatisfaction and a decrease in the overall sale of life insurance policies.4 As a result, “[f]ewer than half of people in the United States aged twenty-five to sixty-four have life insurance coverage.”5
To combat these issues, the life insurance industry has started taking steps to improve underwriting speed and efficiency.6 Most notably, the industry has implemented accelerated underwriting, which can eliminate the use of paramedical exams and laboratory testing in underwriting decisions.7 Rather than rely on medical tests, accelerating underwriting facilitates faster determinations on underwriting offers through the use of new data tools and modeling techniques.8 This is accomplished through the utilization of certain digital technology and artificial intelligence known as “insuretech.”9 While originally used simply to remove barriers to insurance purchasing, insuretech has expanded to making risk selection decisions easier for insurers.10
Insuretech Utilized in Accelerated Underwriting
Insuretech encompasses all technological advancements that are implemented to generate efficiencies in the insurance industry.11 Thus, insuretech includes “smartphone apps, consumer activity wearables, claim acceleration tools, individual consumer risk development systems, online policy handling, [and] automated compliance processing.”12 Social media is also considered to be a subset of insuretech, as it provides information on an applicant’s interests and general life style activities.13 Beyond actual software products, the term insuretech also encompasses start-up companies creating technological advancements in the insurance industry.14 These companies focus mainly on the distribution channel to create efficiencies in the insurance purchasing process.15
A significant asset of insuretech is the innovative and efficient way it accumulates relevant information on a life insurance applicant. Specifically, activity wearables, also referred to as “Internet of Things” devices, collect information on physical activity, heart rate, and personal habits.16 In addition, insuretech allows insurers to quickly gather consumer data such as prescription medications, credit history, and motor-vehicle information from third-party sources.17 Insuretech is also utilized to obtain an applicant’s purchase history, as well as collect behavioral data through cell phone usage and social media activity.18 This ability to receive a constant stream of data allows for “continuous underwriting,” which allows for smart contracts written in code to change in conjunction with any change in the insured’s risk profile or health status.19
The use of insuretech expands beyond mere data collection, however, by additionally providing statistical algorithms and rules engines to assist in underwriting decisions.20 This benefit is obtained through the use of a form of artificial intelligence known as “machine learning systems.”21 This form of artificial intelligence creates analytical models from obtained applicant data to identify norms of human behavior and find-risk correlating patterns.22 Machine learning systems use these norms to determine mortality risks that may otherwise be undetected by human underwriters.23 Moreover, machine learning systems use collected data to effectively calculate an insurance applicant’s “customer lifetime value,” which is a metric that formulates the difference between the revenue an insurer is likely to gain from an applicant and the insurer’s expenditures over the lifetime of the insurance relationship.24
Perceived Benefits and Potential Drawbacks of Accelerated Underwriting
As accelerated underwriting practices have only recently begun to develop, their full impact in the insurance industry has not been determined. Nevertheless, some general benefits and perceived drawbacks on the current implementation of accelerated underwriting through the utilization of insuretech have begun to emerge. Provided below is a non-exhaustive list of these perceived benefits and potential drawbacks on the practice of accelerated underwriting.
At bottom, accelerated underwriting is utilized for the purpose of increasing the speed of underwriting decisions.25 This increased speed is created by insuretech gathering personal and behavioral data on applicants almost instantaneously through third party sources.26 In addition, insuretech’s ability to obtain an applicant’s medical data alleviates the need for an applicant to complete a medical exam for an insurer, which was generally the most time intensive aspect in traditional underwriting practices.27 Lastly, machine learning systems can generally make quicker risk assessments through the use of statistical algorithms and rules engines, and can also generally move claims through the system at a faster pace.28 Lastly, the digitization of the application process through insuretech has made the application process much faster for applicants to complete.29
The increased operational speed resulting from the use of insuretech has also reduced the costs generally associated with insurance underwriting.30 This cost reduction occurs by insuretech either automating or outright eliminating certain routine manual tasks.31 In addition, insuretech has reduced costs through more efficient data collection, eliminating time intensive and expensive efforts on the part of human underwriters to acquire the information necessary for underwriting decisions.32
There is a general belief that insuretech’s ability to analyze a significantly greater amount of data is also likely to increase the accuracy of underwriting determinations.33 This belief is based primarily on the fact that insuretech can perform continuous data collection, making it possible for insurance companies to revise insurance contracts as new risk factors appear.34 Moreover, this continuous data stream will potentially allow insurers to identify adverse events before they occur.35 In essence, therefore, the accuracy of insuretech may “take underwriting from a detectand-repair mindset to a predict-and-prevent philosophy.”36
Conversely, however, there is also a viewpoint that the non-reliance on medical exams in the accelerated writing process effectively sacrifices comprehensiveness in favor of speed and cost reduction.37 Moreover, assuming arguendo that reliance on insuretech does facilitate more accurate risk assessments, such accuracy may not ultimately be a positive for the life insurance industry for the reasons addressed more fully infra. It is therefore unclear whether risk assessment accuracy is a true benefit of accelerated underwriting.
Accelerated underwriting is viewed as more likely to appeal to younger generations, most notably millennials.38 Specifically, younger generations are more likely to seek out life insurance if the application process is streamlined through digital applications.39 Additionally, accelerated underwriting can potentially offer younger applicants more personalized insurance plans that allow for lower premium payments.40 This is particularly important for younger applicants that may otherwise be unwilling to take on the costs of a life insurance policy due to increasingly higher student loan debts and other financial obligations.41 Such personalized policies may allow for insurers to increase their business with younger and middleincome markets, which are generally necessary to subsidize insurance policies to applicants with higher mortality risks.42
Nevertheless, the increased personalization of insurance plans may not benefit the insurance industry as a whole. As discussed more fully infra, the increased personalization of insurance policies may reduce or otherwise eliminate risk pools that are necessary to sustain the insurance industry.
The reliance on insuretech in accelerated underwriting may eliminate bias that is otherwise potentially present in human underwriting.43 Such bias removal would occur through underwriting determinations made pursuant to statistical algorithms and rules engines rather than subjective evaluations of certain risk factors.44
While seemingly eliminating bias through the removal of subjective underwriting determinations, bias may potentially be present in certain algorithms used by insuretech. Specifically, insuretech may evaluate certain variables such as socio-economic status, educational attainment, genetic predisposition to diseases, credit history, and living areas that may have a discriminatory impact on certain classes of insurance applicants.45 Moreover, human intervention will still likely be necessary to monitor and revise the rules running utilized insuretech,46 thereby presenting the possibility that subjective human biases may be implemented into algorithms utilized for underwriting decisions.
Insuretech can also potentially minimize insurers’ risk of financial exposure by identifying major mortality risks of insureds before they occur. This is true through the utilization of insuretech that constantly monitors both insureds’ health status through wearable devices47 and personal behavioral data through call records and social media activity.48 Put simply, constant monitoring may allow insurers to adjust policy premiums based on new mortality risks, thereby minimizing the risks that insurers will miscalculate an insured’s customer lifetime value.
Unlike most industries, the insurance industry is mostly regulated by state law.49 As a result, producers and users of insuretech will have a difficult time creating nationwide underwriting practices.50 This is particularly true in the context of the procurement of genetic data, the regulation of which varies greatly from state to state.51 Moreover, the ongoing issue of compliance will create significant costs and time-consuming efforts for producers of insuretech, particularly for start-up companies.52 As a result, it is unclear whether nationwide implementation of insuretech is practically and financially feasible.
The increased digitalization of insurance applications makes insurers more susceptible to fraudulent procurements of insurance.53 This is true because potential fraudsters have anonymity in the increasingly digitized insurance application process, and are also able to obtain personal information of other individuals through the same methods utilized by insuretech.54 In addition, the lack of human interaction makes it harder for insurers to verify the identity of applicants and the risks they are looking to underwrite.55
Despite these concerns, there is also a belief that insuretech can actually be utilized to prevent fraud.56 Specifically, insuretech may be able to create and utilize fraud-detection models such as natural language processing programs that determine truthful expressions through analysis of normal speech patterns.57 These models may also be able to spot potential fraud much faster than human underwriters, allowing insurers to take quicker steps to eliminate fraudulent activities.58 Insuretech may also be able to create fraud risk profiles based on social media, public records, and internal data sources that can spot attempts at fraudulent insurance procurement at the outset of the application process.59
The increased access to personal data through insuretech creates privacy concerns for insurers.60 This is particularly true because such privacy regulations are generally regulated by state law61 and, therefore, making it difficult—if not impossible—for insurers to establish nationwide privacy practices. In addition, insurers will likely have significant expenditures to protect against insureds’ private information being improperly revealed or otherwise obtained through third party hacking practices. Lastly, the ability to obtain personal information through insuretech without the insureds’ direct knowledge may create customer trust issues.62
Artificial intelligence systems utilized in insuretech have the potential to become inaccurate or invalid over time due to changes that make their knowledge bases and algorithms obsolete.63 As a result, constant maintenance of insuretech is necessary to ensure configuration and operating engines do not become sub-optimal.64 Such maintenance will create a substantial cost to insurers.65 Moreover, insurers will need to incur a significant amount of expenses on mission-critical software to ensure continuous operations of insuretech or, in the event a malfunction or software failure occurs, to ensure fast recovery of stored information.66
Accelerated underwriting is still generally limited to smaller insurance policy amounts.67 In fact, accelerated underwriting generally sets maximum issue ages and other eligibility requirements to control risk selection.68 This ensures that accelerated underwriting is only utilized for more straightforward risk assessments, leaving more complex cases to human underwriters where expert human insight is particularly valued.69 As a result, the costs and risks associated with accelerated underwriting may not ultimately be beneficial to insurers when compared to its limited scope and overall application.
Current Issues with the Utilization of Accelerated Underwriting
The practical application of accelerated underwriting is the most immediate concern to insurers. This is particularly true in the context of post-claim underwriting and the reliance on e-signatures, two areas that are impacted by the implementation of accelerated underwriting. Below, both areas of concern are addressed.
The general conflict between the justifications for post-claim underwriting and the practices of accelerated underwriting.
Post-claim underwriting occurs when an insurer waits until after a claim is made to determine whether the claimant/beneficiary is eligible for the insurance coverage previously provided.70 This practice is generally criticized as an improper delegation of the insurer’s obligation to perform thorough underwriting investigations during the initial application process.71 Put simply, the practice of post-claim underwriting is viewed by some as a means for insurers to rescind policies that were knowingly issued without sufficient investigation for the purpose of collecting immediate profits on premium payments.72 As a result, many states prohibit post-claim underwriting practices.73
Nevertheless, as an exception to this general restriction, post-claim underwriting is typically allowed when a material misrepresentation is made by an applicant during the application process.74 This is particularly true in the context of life insurance policies “in the event that a policy holder dies within their life insurance policy’s contestable period.”75 This exception is justified on the basis that a death within the contestable period of an insurance policy creates a reasonable suspicion that a misrepresentation was made during the application process.76 Thus, a life insurance policy may generally be rescinded when:
The applicant made a misrepresentation during the application process;
The misrepresentation was material; and
The applicant was aware he made a misrepresentation.77
Going forward, however, it may be more difficult to determine when a misrepresentation sufficient to rescind a life insurance policy is made in the context of the accelerated underwriting process. This is true because of the dueling principles that determine whether rescission on the basis of applicant misrepresentations is justified. The first of these principles is that insurers are generally able to “rely upon the insured to provide such information as it needs to determine whether to provide coverage.”78 In essence, this principle is that life insurance applicants must fully disclose all necessary information at the inception of the application process. Id. With accelerated underwriting, however, insurers are able to independently accumulate personal data on applicants, thereby cutting down the number of questions asked of applicants during the application process.79 As a result, it will be more difficult to discern when an insured had knowledge of a “misrepresentation” made during the application process that impacted the issuance decision.
The second principle is that an insurer’s right to material facts may be waived “by neglect to make inquiries as to such facts, where they are distinctly implied in other facts of which information is communicated.”80 Of course, by outsourcing applicant data collection to insuretech, insurers are less involved in the initial application process, thereby providing less opportunity to make additional inquiries of the applicant. Put simply, an insurer’s ability to know what additional factual inquiries should be made is limited in the “one size fits all” approach of insuretech data collection.81 It is therefore unclear what investigative burdens will be placed on insurers who rely on insuretech to obtain the bulk of the information necessary for underwriting decisions.
Conflicting standards on post-claim underwriting potentially applicable to accelerated underwriting practices.
While not addressing the impact of insuretech directly, a circular letter issued by the New York Department of Financial Services on January 26, 2017 (the “Circular Letter”) provides insight on the potential ability to rescind life insurance contracts procured through accelerated underwriting in the future.82 In the Circular Letter, the New York Department of Financial Services provided that a life insurance provider can only demand a claimant/beneficiary comply with all aspects of the post-claim underwriting process when it has “actual proof of a material misrepresentation within two years of the policy’s date of issue or the effective date of the increase or change.”83 This decision was based on the recent practice of insurers placing “the burden on the beneficiary to obtain the deceased insured’s medical records.”84 Such conduct was considered improper in part because such beneficiaries “may have no legal standing to waive the deceased insured’s physician-patient privilege and obtain the necessary medical records.”85 Thus, the New York Department of Financial Services determined it was improper for insurers—absent evidence of a material representation—to require medical records that were likely unattainable in conducting post-claim underwriting when such records were not deemed necessary in the initial application process.86 Of course, the utilization of accelerated underwriting will increase insurers’ issuance of life insurance policies without requiring applicants to submit medical records in conjunction with their applications. In essence, therefore, the Circular Letter restricts insurers’ ability to rescind life insurance contracts due to the use of underwriting practices generally utilized in accelerated underwriting.
Importantly, however, the post-claim underwriting standard advocated in the Circular Letter is not consistent with federal court decisions Tran v. Kansas City Life Ins. Co.87 and Gary v. USAA Life Ins. Co.,88 which were issued around the same time period.89 In Tran, the beneficiary argued that the insurer’s claim denial due to the insured’s failure to disclose his diabetes and high blood pressure diagnoses in the application process resulted from improper post-claim underwriting because it was based on “medical records, it acquired after the policy became effective.”90 The Central District of California rejected this argument simply because the insured died during the underlying policy’s contestable period.91 The Court further rejected the beneficiary’s argument that the insurer failed to adequately investigate the insured’s medical records during the initial underwriting process because “an insurer may rescind an insurance contract for material misrepresentations even if the insurer issued the policy without investigating the applicant’s medical history.”92 Lastly, the Court also rejected the argument that the insured’s failure to disclose his diabetes did not constitute a misrepresentation because the insured may have been confused by the application only requiring the disclosure of any “disease or disorder,” which does not technically encompass diabetes.93 This argument was rejected on the basis that a misrepresentation sufficient to justify post-claim underwriting does not need to be “intentional,” and that the application was not confusing.94
In Gary, the United States District Court for the District of Maryland reached a similar result in granting summary judgment to an insurer that denied a life insurance claim pursuant to post-claim underwriting.95 Specifically, the insurer argued that the insured’s failure to disclose an echocardiogram constituted a misrepresentation that justified rescission of the underlying policy.96 The beneficiary argued that the insurer’s knowledge that the insured suffered from a heart murmur “without requesting her medical records means that the failure to disclose the echocardiogram was not material because the insurer already had a reason to request medical records (to learn more about the murmur).”97 The Court rejected this argument on the grounds that the insurer had reason to believe the heart murmur was benign and, therefore, did not have to request the insured’s medical records in conjunction with the underlying application.98 In essence, therefore, the Court in Gary also held that an insurer can rely on medical records in post-claim underwriting that were neither obtained nor requested in the original underwriting process.
This apparent conflict on the scope of valid post-claim underwriting is further evidenced by recent evaluations of statutory law governing insurance contracts in the states of Tennessee and South Carolina. Specifically, the South Carolina Department of Insurance (the “SCDI”) issued a bulletin on April 3, 2019 regarding compliance with South Carolina Code, Section 38-63-220.99 This statute provides that each insurance contract must contain a provision that states the underlying policy is “incontestable as to the truth of the application for insurance” once it has been in force “for a period of two years from their date of issue.”100 In its April 3, 2019 bulletin, the SCDI clarified that this language establishes that “the South Carolina General Assembly mandated that if an insurer want[s] to challenge the truthfulness of the application for insurance, it must do so during the first two years of the policy.”101 As a result, no insurance policy governed by South Carolina law can be challenged on the basis of a misrepresentation in the application process if the underlying policy has been in effect for two years. This determination is consistent with the Circular Letter by generally prohibiting an underwriter’s ability to demand a claimant/beneficiary comply with post-claim underwriting investigations. More importantly, the rationale for both the determination of the SCDI and the Circular Letter is the same. Specifically, both the SCDI and the Circular Letter articulate that an insurer should engage in underwriting investigations that extend beyond the practices of accelerated underwriting during the application phase, instead of waiting for a claim to trigger more traditional investigation processes.
Conversely, on April 9, 2019, the Tennessee House of Representatives proposed adding the following language to Tennessee Code Annotated, Section 56-7-102, which concerns the general construction and interpretation of insurance contracts:
An insurance company may determine its obligations under a policy of insurance as to any and all parties or claimants through a declaratory judgment action, an interpleader claim or action, or both. The filing of such action or claim creates a rebuttable presumption the insurance company is acting in good faith when making a determination of its obligations under a policy of insurance.102
This proposed amendment would allow an insurer to engage in post-claim underwriting even when there is no apparent evidence a misrepresentation was made during the application process. Moreover, this proposed amendment would not put any time limitations on an insurers’ ability to perform post-claim underwriting. In essence, therefore, this proposed amendment to Tennessee statutory law is in line with the holdings of Tran and Gary. Specifically, all three authorities would provide support for an insurer to engage in post-claim underwriting even after utilizing accelerated underwriting practices in the initial application process.
Based on the conflict between the holdings in Tran and Gary and the Tennessee House of Representative’s proposed amendment to statutory law on one hand, and the Circular Letter and the SCDI on the other, it remains unclear what impact accelerated underwriting practices will have on post-claim underwriting. At a minimum, however, these conflicting opinions illustrate that new challenges to post-credit underwriting are likely to occur based on the practices utilized in accelerated underwriting.
Reliance on E-Signatures
General benefits and concerns of e-signatures.
Insurers are more reliant on e-signatures in accelerated underwriting practices that utilize digitized insurance applications and smart contracts. This reliance on e-signatures is generally viewed as a positive, as it allows insurers to reduce cycle times.103 In addition, by eliminating the need for clients to sign contracts in person, smart contracts allow insurers to engage in continuous underwriting.104 This is true because insures can sign automatically when changes are made to their insurance policies, thereby making such changes effective almost instantaneously.105
Despite these benefits, however, the increased utilization of e-signatures has also created practical concerns for insurers. Most notably, the increased use of e-signatures has left insurers more susceptible to fraud in the insurance application process.106 As discussed above, this is true because fraudulent insurance practices thrive on the anonymity achievable through e-signature authorizations.107 Fraudsters can therefore potentially commit the following acts through the use of e-signatures:
Obtain insurance policies for fictitious beneficiaries;
Open and subsequently cancel an insurance policy for the benefit of an insurance agent; and
Modify insurance information in order to reduce premium payments owed thereunder.108
Legal and practical concerns on e-signature verification.
Going forward, the insurance industry will need to determine the best practices for preventing fraud through the utilization of e-signatures. This is most likely achieved through authentication processes that confirm the identify of a party signing a document through an e-signature.109 This may prove difficult, however, because the laws addressing the level of authentication necessary to validate an e-signature differs between states.110 As a result of these differing standards, it will likely be difficult for insurers to create a nationwide authentication process for e-signatures.
Beyond the legal issues, there are also practical considerations that insurers must contemplate when addressing this issue of authentication. Specifically, insurers could implement additional digital devices that help identify applicants and insureds utilizing e-signatures through multiple data sources. In addition, insurers could require applicants and insureds to go through more extensive verification processes, such as providing multiple forms of personal identifying information like checking account numbers and social security numbers.111 These numbers can then be analyzed by insuretech to ensure they match records from credit bureaus and other information sources.112 Of course, such practices may ultimately eliminate the efficiency and client convenience that accelerated underwriting is designed to facilitate. Additionally, a more extensive authentication process may create client mistrust on whether information provided to insurers is adequately protected.113
In summary, it is currently unclear whether the reliance on e-signatures is the accelerated underwriting process is ultimately beneficial to the insurance industry. It is also unclear whether potential issues with e-signatures can be reduced or eliminated in a way that is both cost effective and customer friendly. These issues will need to be resolved in order to determine the true benefit of accelerated underwriting practices.
Overarching Industry Concerns Presented by Accelerated Underwriting
Beyond the immediate practical concerns, the implementation of accelerated underwriting creates broader issues that the industry should be cognizant of going forward. One such issue is the overarching impact accelerated underwriting and the implementation of insuretech will have on the insurance job market. In addition, the utilization of accelerated underwriting may have a significant impact on the formation and sustainability of risk pools generally utilized in insurance underwriting. These issues should be a key concern to the industry going forward, as they will ultimately impact the overall viability of the insurance industry.
Accelerated Underwriting’s Impact on the Insurance Industry Job Market
Potential risk of job reduction and job loss.
The growing use of automation machines and artificial intelligence has had a significant impact on the overall job market in the United States.114 This is true because announced technologies can complete 30% or more of the vocational tasks associated with roughly 60% of jobs in the United States.115 The insurance industry as a whole certainly falls within this range, as announced technologies are potentially capable of performing around 65% of insurance agents’ job-related tasks.116 In addition, announced technologies are estimated as being able to complete roughly 35% of the job tasks generally performed by insurance underwriters.117 As a result, a study performed by CareerCast.com in 2015 determined that insurance underwriters were “No. 9 of 10 endangered jobs.”118 This determination was made on the basis that “[s]treamlined processes allow agents to take on the work previously handled by underwriters.”119 Consistent with this determination, the U.S. Bureau of Labor Statistics has projected that between 2012 and 2022 underwriting jobs throughout all industries will shrink from 106,300 to 99,800.120
Undoubtedly, the rise in insurance underwriter job tasks that can potentially be automated is the result of the creation and utilization of insuretech. Nevertheless, there is still a general consensus that human underwriters cannot be replaced completely by artificial intelligence, particularly in more complex cases.121 Moreover, underwriters will still likely be needed to monitor and revise the rules running utilized insuretech.122 While this need for human reasoning is acknowledged despite the use of insuretech, it remains unclear the exact role human underwriters will have as the insurance industry expands its use of insuretech.123
Ideally, insuretech will merely augment human capabilities, allowing human underwriters to work alongside artificial intelligence to further enhance the underwriting process.124 This is particularly true given the fact that it is still considered unlikely a single artificial intelligence could fully underwrite facultative insurance applications.125 Thus, the job duties of a human underwriter may change towards the utilization of skill sets that incorporate or otherwise augment the performance of insuretech in the underwriting process.126 What these job duties will be, however, as well as how these new job duties will change insurance underwriters day-to-day workflow— and by extension their net income—remains undetermined.
Potential liability issues created by the utilization of insuretech.
In considering what role insuretech will play in the underwriting process, it is important to be mindful of the potential liability issues that may result from the utilization of insuretch. Specifically, liability issues could arise in the event underwriting determinations are made pursuant to statistical algorithms used by insuretech that are deemed to be incorrect, defective, or otherwise discriminatory. Liability issues may also arise if data collection techniques performed by insuretech are deemed to violate either applicable privacy laws or state insurance regulations. At the moment, there is very little regulation on the utilization of artificial intelligence in general,127 let alone within the insurance industry. As a result, it is unclear “who is liable for artificial intelligence mistakes.”128 This lack of regulation creates undetermined legal risks to insurance companies that utilize insuretech. Specifically, it is unclear whether the provider of insuretech would be wholly liable for the mistakes of its offered products, or if insurers would be subject to liability for their reliance on such technology. It is also unclear what responsibilities a user of insuretech will have to ensure the accuracy and proper functionality of its utilized software. Compounding these issues is the concern that artificial intelligence may be a “legal person” capable of being sued,129 thereby potentially making a user of such technology jointly and severally liable for its mistakes under principles of agency law.
These issues of insurer liability are enhanced when a human underwriter works alongside insuretech. Insurance underwriters are undoubtedly employees of insurance companies.130 Insurance companies are therefore liable for the acts of underwriters performed within the scope of their employment.131 As a result, life insurance companies generally implement rules that help “their agents and underwriters determine which applicants to accept and which to reject.”132 Such rules are not as well defined, however, with respect to reliance on insuretech. Therefore, from a liability standpoint, it is unclear how much an insurance underwriter can rely on insuretech “within the scope of employment.” This lack of guidance on the interplay between human underwriters and insuretech could create issues if underwriters merely follow the recommendations of algorithms and/or data collections provided by insuretech that are determined to be faulty or have a discriminatory impact. Conversely, however, liability issues may also arise if an underwriter deviates from an insuretech recommendation, as such a deviation could be used as evidence that the underwriting decision was the result of improper bias. This result would effectively dis-incentivize the use of human intervention in the underwriting process, thereby making the role of human underwriters more vulnerable to reduction. Until these issues of liability are resolved, it is unclear what the full legal impact of the utilization of insuretech will be.
What level of human involvement in accelerated underwriting is beneficial?
The potential issues on the insurance industry market noted above highlight the more general question of how much insurance companies will want to rely on insuretech in the future. Human insight is undoubtedly an irreplaceable asset that can successfully augment the potential advantages insuretech provides to underwriting practices.133 Nevertheless, the desire to further expand the use of insuretech in the underwriting process remains present.134 Therefore, the insurance industry will have to determine how much of the human element it is willing to sacrifice in the underwriting process for the sake of speed and efficiency.
Accelerated Underwriting’s Impact on Risk Pools
Potential reduction of risk pools.
“Risk pooling” is the traditional underwriting system utilized by insurance companies.135 Under this system, insurers produce an approximate risk profile for each insurance applicant.136 Insurance companies then group the individual applicants into risk pools where premium payments are grouped together, effectively mitigating the risk of miscalculations or inaccuracies in the underwriting process equitably across the entire pool.137 This practice allows insurers to accept similar infrequent risks from a significant number of applicants, thereby reducing its overall financial risk in the event a claim must be paid.138
The utilization of accelerated underwriting through insuretech, however, may lead to the reduction of the use of risk pools, or otherwise reduce the size of utilized risk pools, in favor of more personalized insurance policies.139 This is true because data obtained from wearable devices, smart phones, and social media accounts allow insurers to obtain more individualized data for each applicant.140 Moreover, machine learning systems allow insurers to create more personalized insurance experiences through the use of calculations on behavioral trends unique to each individual applicant.141 This information will potentially allow insurers to evaluate an applicant’s risks, and by extension their corresponding premium charges, more accurately.142 This will effectively minimize the need for one insurer to subsidize the insurance of another.143 Such a result will ultimately reduce the size of risk pools, if not eliminate them entirely.144
Potential drawbacks of personalized insurance plans.
In many areas of the insurance industry, more personalized insurance plans are seen as a benefit to insurance applicants.145 This viewpoint may be shortsighted, however, as more personalized insurance plans may “also create corresponding harms to customers.”146 Specifically, individuals with distinct health risks that would otherwise struggle to obtain insurance benefit from the use of large risk pools.147 This is true because an insurer’s risk in insuring such individuals is reduced in a large risk pool that encompasses many individuals with far less insurance risks.148 As a result, individuals with “with high risk factors that normally would balance out as part of a risk pool may instead lose their subsidy.”149
Even if they are not denied insurance, applicants can be harmed by the reduction of risk pools through significantly higher premium payments.150 Through the underwriting process, an insured’s insurance premium is determined based on his or her mortality risks in comparison to the risks of the average insured party.151 Obviously, therefore, an individual with higher mortality risks generally pays a higher premium for life insurance.152 Such premiums of high risk individuals are still able to be reasonably affordable, however, when they are placed in a large risk pool with a significant number of individuals with low premiums.153 If such high risk pools are eliminated or substantially reduced, no such offset is possible.154 As a result, the premiums charged to high risk individuals will rise significantly.155 Such a rise in premiums could effectively price out high risk individuals from obtaining insurance.
What level of risk assessment accuracy is beneficial to the insurance industry?
The issues potentially presented by the reduction of risk pools present a larger question on whether complete underwriting accuracy is an ultimate benefit to the insurance industry. In general, an individual’s need for insurance corresponds with their likelihood to procure insurance.156 Thus, while younger individuals are the most likely to benefit from more personalized insurance,157 they are also the least likely to obtain health insurance based on their health, age, and limited financial and familial obligations. This issue of procuring business from younger, healthier individuals that effectively subsidize higher risk individuals that want insurance will be compounded if the predictions of mortality approach perfection. Put simply, healthy and low risk individuals would be less likely to procure insurance if life expectancy predictions could be made with near certainty, effectively confirming that they do not have a significant need for life insurance. Moreover, even if such individuals do procure insurance, their premiums could potentially be so low that they would not be able to subsidize the insurance premiums of less healthy individuals. Such a result could effectively create a scenario where the individuals with the greatest need for insurance would have to be charged exorbitant premiums to keep the insurance industry functional. The insurance industry will need to consider these issues as it continues to utilize insuretech in the underwriting process.
The overall impact of accelerated underwriting and the utilization of insuretech on the life insurance industry is unknown. As the industry moves forward, it should consider both the immediate and long-term implications of its utilization of accelerated underwriting practices. Proper assessment of the issues and business practices will allow the industry to determine the ultimate benefits of accelerated underwriting, as well as the most effective use of insuretech in the improvement of underwriting practices in general.
*The authors would like to thank their colleague Heather Hall for her research contribution to this paper.
1 Noorhannah Boodhun & Manoj Jayabalan, Risk prediction in life insurance industry using supervised learning algorithms, 4 Complex & Intelligent Systems 145, 146 (2018).
2 Daniel Kurt, 7 Factors That Affect Your Life Insurance Quote, INVESTOPEDIA (last updated Oct. 9, 2018), https://www.investopedia.com/articles/investing/102914/7-factors-affect-....
3 Boodhun & Jayabalan, supra note 1, at 146.
5 Sam Lewis, Student Author, INSURETECH: AN INDUSTRY RIPE FOR DISRUPTION, 1 Geo. L. Tech. Rev. 491, 491 (2017).
6 Greg Iacurci, Technology is streamlining the process of issuing life insurance policies, InvestmentNews (Jan. 10, 2018, 2:37 PM), https://www.investmentnews.com/article/20180110/ FREE/180119992/technology-is-streamlining-the-process-of-issuing-life-insurance.
7 Dianne Schuetz, et al, BUSTED: TOP FIVE MYTHS ABOUT ACCELERATED UNDERWRITING, 34 ON THE RISK 32, 32 (2018).
8 Id.; Iacurci, supra note 6.
9 Patrick Wraight, A Quick (No Really, Super Quick) Insuretech Primer, Insurance Journal (June 13, 2018), https://www.insurancejournal.com/blogs/academy-journal/2018/06/13/491903....
12 Ben Deda, What is InsurTech and how can you harness its disruptive powers?, Vertafore (Sept. 20, 2017), https://www.vertafore.com/resources/blog-posts/what-insurtech-and-how-ca....
13 Lewis, supra note 5, at 495.
14 Wraight, supra note 9.
16 Lewis, supra note 5, at 493.
17 Iacurci, supra note 6.
18 Tyler Tappendorf, Five InsureTech Trends and What They Mean for Microinsurance, FinDev Gateway (Feb. 2017), https://www.findevgateway.org/blog/2017/feb/five-insuretech-trends-andwh....
19 Lewis, supra note 5, at 496.
20 Iacurci, supra note 6.
21 Lewis, supra note 5, at 494.
24 Mateusz Hapon, 6 Ways Machine Learning is Changing Insurance, Netguru (Dec. 13, 2018), https://www.netguru.com/blog/machine-learning-insurance-insurtech-fintec....
25 Iacurci, supra note 6.
26 Id.; Tappendorf, supra note 18.
27 Iacurci, supra note 6.
28 Id.; Hapon, supra note 24.
29 Iacurci, supra note 6.
30 Timothy L. Rozar, Life Insurance in a Time of Rapid Technological Change, RGA (Apr. 1, 2015), https://www.rgare.com/knowledge-center/media/articles/life-insurance-in-....
33 Jared Shelly, Here’s How Artificial Intelligence Is Poised to Transform Insurance Underwriting for the Better, Risk & Insurance (Jan. 28, 2019), https://riskandinsurance.com/aitransforms-underwriting. 34 Lewis, supra note 5, at 496.
36 Shelly, supra note 30.
37 Lewis, supra note 5, at 493
38 Thiru Sivasubramanian, How Insuretech Can Help Millennials Buy Life Insurance, Forbes (Nov. 20, 2018 8:15 AM), https://www.forbes.com/sites/forbestechcouncil/2018/11/20/howinsuretech-....
40 Id.; Rozar, supra note 30.
41 Sivasubramanian, supra note 35.
42 Rozar, supra note 30.
43 Lewis, supra note 5, at 494.
45 Id. at 496; NY Circular Letter No. 2019-1 (Jan. 18, 2019), 2019 WL 271541 at *2-4 (stating the use of external data sources, algorithms, and predictive models in the life insurance underwriting process presents the “immediate concern” of potential unlawful discrimination).
46 Susanne Sclafane, Tasks Done by Agents, Underwriters, CEOs Can Be Automated: McKinsey, Carrier Management (Dec. 21, 2015), https://www.carriermanagement.com/ features/2015/12/21/149066.htm.
47 Lewis, supra note 5, at 495.
48 Tappendorf, supra note 18.
49 Lewis, supra note 5, at 498.
51 Lawrence O. Gostin & James G. Hodge, Jr., GENETIC PRIVACY AND THE LAW: AN END TO GENETICS EXCEPTIONALISM, 40 Jurimetrics J. 21, 22 (1999).
52 Lewis, supra note 5, at 498.
53 Fighting Insurance Application Fraud, SAS (last visited Apr. 14, 2019), https://www.sas.com/ en_nz/whitepapers/insurance-application-fraud-109925.html.
56 Lewis, supra note 5, at 497; Hapon, supra note 24; Rozar, supra note 30.
57 Lewis, supra note 5, at 497.
58 Hapon, supra note 24.
59 Rozar, supra note 30.
61 Lewis, supra note 5, at 498.
62 Rozar, supra note 30.
63 Kareem S. Aggour, et al, Automating the Underwriting of Insurance Applications, 27 AI Magazine 36, 46-47 (2006).
65 Id. at 47.
67 Iacurci, supra note 6.
68 Schuetz, et al, supra note 7, at 32.
69 Id. at 34.
70 Lewis v. Equity Nat’l Life Ins. Co., 637 So.2d 183, 186 (Miss. 1994).
71 How to Recognize Post-Claim Underwriting, https://law.freeadvice.com/insurance_law/ insurers_bad_faith/recoginize-post-claim-underwriting.htm (last visited Apr. 14, 2019).
72 Post-Claim Underwriting: How Insurance Companies Plan Not to Pay You, Dolman Law Group (July 28, 2017), https://www.dolmanlaw.com/post-claim-underwriting.
73 See, e.g., 28 Tex. Admin. Code § 3.3823 (stating that certain life insurance policies “shall be utilized in such manner as would result in post-claims underwriting.”); White v. Cont’l Gen. Ins. Co., 831 F. Supp. 1545, 1562 (D. Wyo. 1993) (holding allegations of post-claim underwriting presented a question of fact as to whether insurer acted in bad faith); 31 Pa. Code § 89a.110 (the model regulation on the general restriction of post-underwriting long-term care insurance contracts).
74 See N.Y. INS § 3105(b)(1) (stating a material representation in the insurance application process allows an insurer to avoid or defeat recovery of a claim under the insurance contract).
75 Tran v. Kansas City Life Ins. Co., 228 F. Supp. 3d 1068, 1078 (C.D. Cal. 2017).
76 See Yang v. Peoples Benefit Ins. Co., No. CIV F 06-458 AWI DLB, 2007 WL 1555749, at *2 (E.D. Cal. 2007) (stating that underwriting reviews of life insurance policies when the insured dies within the contestable period is standard practice in the life insurance industry).
77 Tran, 228 F. Supp. 3d at 1074; N.Y. Ins. § 3105(b)(1) (“No misrepresentation shall avoid any contract of insurance or defeat recovery thereunder unless such misrepresentation was material. No misrepresentation shall be deemed material unless knowledge by the insurer of the facts misrepresented would have led to a refusal by the insurer to make such contract.”).
78 Bhakta v. Hartford Life & Annuity Ins. Co., 673 F. App’x 762, 764 (9th Cir. 2016).
79 Iacurci, supra note 6.
80 Bhakta, 673 at 764 (quoting Cal. Ins. Code § 336).
81 Schuetz, et al, supra note 7, at 34.
82 NY Circular Letter No. 2017-1 (Jan. 26, 2017), 2017 WL 388170 at *1-3.
83 Id. at *1.
85 Id. at *2.
86 Id. at *1-3.
87 228 F. Supp. 3d 1068 (C.D. Cal. 2017).
88 229 F. Supp.3d 365 (D. Md. 2017).
89 Tran was issued on January 5, 2017 and Gary was issued on January 17, 2017.
90 228 F. Supp. 3d at 1078.
92 Id. (citation omitted).
93 Id. at 1076.
95 229 F. Supp. 3d at 381.
97 Id. at 380-81.
98 Id. at 381.
99 SC Bulletin No. 2-2019 (Apr. 3, 2019), 2019 WL 1498182, at *1-2.
100 SC ST § 38-63-220(d).
101 2019 WL 14988182, at *1.
102 2019 TN H.B. 348 (Jan. 30, 2019).
103 Hapon, supra note 24.
104 Lewis, supra note 5, at 496-97.
105 Id. at 497.
106 Supra note 53.
110 Maureen Mineham, Are your e-signatures enforceable?, 311 Employment Law Counselor NL 1, 1 (2016).
111 Dawn Lewallen, TECHNOLOGY: LEGAL CONSIDERATIONS IN ELECTRONIC SIGNATURE INTEGRITY, Advanced Real Est. Drafting 9.IV (2018).
112 Supra note 53.
113 Rozar, supra note 30.
114 Sclafane, supra note 36.
120 John Najarian, Is Automation Going to Put an End to the Underwriting Profession, Genre (Dec. 13, 2015), http://www.genre.com/knowledge/blog/is-automation-going-to-endunderwriti....
121 Schuetz, et al, supra note 7, at 34; Shelly, supra note 30.
122 Sclafane, supra note 36.
123 Shelly, supra note 30.
124 Schuetz, et al, supra note 7, at 34.
125 Dihui Lai, Wired to Underwrite: Artificial Intelligence and Underwriting, RGA (Oct. 30, 2018), https://www.rgare.com/knowledge-center/media/articles/wired-to-underwrit....
127 Alexandra M. Jones, Student Author, OLD DAYS ARE DEAD AND GONE: ESTATE PLANNING MUST KEEP ITS HEAD ABOVE WATER WITH THE CHANGING TIDE OF TECHNOLOGY, 11 Est. Plan. & Community Prop. L. J. 161, 180 (2018).
128 Id. at 162.
129 Id. at 167.
130 See Agency Solutions.Com, LLC v. TriZetto Grp., Inc., 819 F. Supp. 2d 1001, 1021 (E.D. Cal. 2011) (“According to Black’s Law Dictionary an ‘insurance underwriter’ is ‘an insurance company employee who is responsible for determining whether to issue a policy and the amount to charge for the coverage provided.’”).
131 See 1 American Law of Torts § 4:3 (articulating that an employer is liable for the acts of its employee while acting within the scope of the employment pursuant to the doctrine of respondeat superior).
132 Brian J. Glenn, THE SHIFTING RHETORIC OF INSURANCE DENIAL, 34 Law & Soc’y Rev. 779, 787 (2000)
133 Schuetz, et al, supra note 7, at 34; Georgia Mosis, Artificial Intelligence and Underwriting: An Augmented Look at the Big Picture, RGA (last visited Apr. 14, 2019), https://www.rgare. com/docs/default-source/newsletters-articles/artificial-intelligence-and-underwriting-v1-final. pdf?sfvrsn=a0a0a688_2.
134 Lai, supra note 125.
135 Lewis, supra note 5, at 495.
136 Id. 137 Id.; Kyle McDonald, The Life Insurance Risk Pool-Part I, Pivot (last visited Apr. 14, 2019), https://www.pivot.com/Learning-Center/Pivot-Blog/Post/1932/The-Life-Insu....
138 McDonald, supra note 137.
139 Lewis, supra note 5, at 495.
140 Id.; Hapon, supra note 24.
141 Lewis, supra note 5, at 495; Hapon, supra note 24.
142 Lewis, supra note 5, at 495.
143 Id. at 495-96.
145 Hapon, supra note 24; Sivasubramanian, supra note 35.
146 Lewis, supra note 5, at 496.
151 Id. at 495-96.
152 Mila Araujo, What Is Insurance Underwriting? Insurance Underwriting Explained, The Balance (last updated Mar. 12, 2019), https://www.thebalance.com/what-is-insuranceunderwriting-2645778.
153 McDonald, supra note 136.
154 Lewis, supra note 5, at 496.
156 Id. at 495-96.
157 Sivasubramanian, supra note 35