Artificial intelligence (AI) is one of the megatrends emerging from the broader digitisation of society and the economy. So far, the discussion about AI has been around for a while, especially the fact that it represents a collection of multiple technologies that enable machines to sense, comprehend, act and learn (Matt Collier,Ricard Fu,Lucy Yin, 2017) so they can perform many of tasks that previously required human effort. Artificial intelligence (AI), deep learning, machine learning and neural networks represent incredibly exciting and powerful machine learning-based techniques used to solve many real-world business problems (Habeeb, 2017).

The combination of AI technology and data analytics is being employed in several companies and industries ranging from space technology, automobile, manufacturing, logistics, financial services, health etc. to automate processes and increase overall efficiency. The benefits promised by AI is already being enjoyed in self-driving cars, advanced robotics, connected machines, business decision, automation, fraud detection, algorithm trading, customer experience, machine diagnosis, drug discovery etc. In medicine, AI today can augment human activity—taking over tasks that range from medical imaging to risk analysis to diagnosing health conditions(Matt Collier,Ricard Fu,Lucy Yin, 2017)

While we rejoice over the immense benefits presented by AI, the adoption of this technology also comes with ethical-moral issues and considerable potential risk. This piece of work aims to discuss the ethical considerations that an Information System professional should consider and address when designing an AI solution to automate health insurer decision to offer someone a new health insurance policy. We shall consider the current application of AI in healthcare insurance and benefits brought about by an automated process and thereafter, we would look at the ethical considerations relating to the adoption of the technology, including the inflexibility of the system to a different scenario and how it affect customers; the doctrine of confidentiality over customer health information and the requirement of utmost professionalism and due care; and, lastly, the owner of the risk in the event of litigation: the insurance company or the software developer.

Health insurance constitute a significant proportion of the healthcare industry and based on data from the Centre for Medical and Medical Services (2019), private health insurance expenditures alone is estimated at $1.1 billion in 2016; this figure represents 34 per cent of the 2016 National Health Expenditure at $3.3 trillion. Health insurance is important because purchasing policy reduces consumers exposure to risks and unpredictable health care expenses. The way health insurance works are that the consumer selects from different grades of health insurance policies, and the insurance company settles each health care expenses incurred by the customer.  An employee may also obtain health insurance package as part of his benefits and as stated in his contract of employment. Management of claims (assessing the authenticity and extent of liability) is cumbersome and fraught with subjectivity and delays, hence the need for automation. 

Although exploration of the possibilities that AI offers in the field of healthcare delivery and management is in its infancy, most notable progress to date is seen in the use of AI for early detection systems supported by algorithms or automated recognition of patterns in patient data. In healthcare Insurance, the technology could bring about effectiveness, cost saving and fraud detection. In Germany, the general cost of inpatient treatment amounts to EUR 73 billion and makes up 30 to 40 per cent of a typical health insurer’s total budget; on average, however, between 8 and 10 per cent of all claims received are incorrect (Mckinsey, 2017). AI offers technology for insurance claim management and can effectively identify and correct wrong claims. The inflexible rule book traditional claim management approach cannot stand the test of time, especially in this digital age; it is obsolete and has been replaced by an intelligent software algorithm that continuously evolves and learns from historical cases. However, while we can look forward to the benefits presented by AI in improving healthcare, the adoption of these technologies is not without considerable potential risks (Hamid, 2016). Hence, the need to have a robust discussion about the morality and ethics surrounding the use of AI for processing customer’s insurance claim.

Naturally, whether humans or corporate organisation, we all hate to be disrupted. With the advent of digital technology that includes AI, probably no industry can immune to the possibility of digital disruption. The matter surrounding the development of machines with the capability to perform the same task as human and even more effectively creates some panic within the corporate environment. Many people have expressed doubts about the possibility of the technology creating unemployment and inequality; some school of thought believes that it could be compromised and the data security is not ascertained; in fact, some have raised the viewpoint of AI bias and racism. All these points of view have come from religious, cultural and ethical perspectives. However, in this section, we would attempt to discuss the ethical consideration that an Information System developer should consider when developing AI for management of healthcare insurance claims. How will cases not previously learnt by the machine not be rejected? Will AI provide an insight into the collective ethical priorities of a different culture that is usually considered when the traditional claim assessment method is adopted? Will the machine ensure that new cases, not previously learnt, are not rejected? Who will be held responsible should there be a lawsuit by a customer?

Consider the “trolley problem” developed by Philippa Foot in 1967: A runaway trolley rushes toward five people tied to the tracks, and it will surely kill them all. Fortunately, you can reach a switch that will turn the trolley onto a side track — but then you notice that one other person is standing there. Is it morally permissible for you to turn the trolley to that side track, where it will kill one person instead of five? Is it not only morally permissible but even morally required (Rolley et al., 2015)?

Image sourced from: Wikipedia

While AI systems may have been trained on comprehensive datasets, in the clinical setting they may encounter data and scenarios that they have not been trained on, potentially making them less accurate and reliable and therefore putting at risk patient safety (Hamid, 2016). But, some right actions may yet have a harmful impact, most often than not, morally acceptable actions are those which side effect was foreseen and considered as unintentional and, maybe, the most appropriate solution is to strike a balance. However, for AI, the problem will probably be, “You are tied to a track/ there is moving trolley and two separate switches. One switch is used by AI, which runs on data and algorithm generated by past health records and present situation. Hence the software is unable to draw from historical data that matches with the new case and information being presented by the customer. As any information that different from the usual poses a challenge to the programmed system, hence, the ethical issue is the possibility overcharge, undercharge or rejection of application for health insurance. These issues call for serious consideration by software developers when considering AI for automation of the claim department. The developer must consider how the machine and human will complement each other, so human reviews that rejected claims before the final decision is passed to the claimant.

Confidentiality is roughly equivalent to the privacy, and there are measures undertaken to ascertain confidentiality are designed to prevent sensitive information from reaching the wrong people while making sure that the right people can, in fact, get it: Access must be restricted to those authorized to view the data in question (Margaret R., 2014). About the health care sector, confidentiality relates to personal information about patients. Problems arise when it comes to deciding how this information should be shared to improve a patient’s care and for better communication amongst care staff.

Similarly, during application for healthcare policy with an insurance company, details of the health condition of the customer are obtained to enable the company to determine the necessary policy to be provided. However, with the advent of Artificial Intelligence, processes for determining the right policy to be issued to the prospective assured, usually done by the technical department, are now being automated. This machine is designed to compare the health condition of a policy seeker with a repository of health cases to determine the right policy. Therefore, it is incumbent on the developer of AI application for an insurance company to ensure that health information provided for the customer is managed in such a way that they are not divulged, manipulated or shared without their consent. The matter of control over customer’s data and information must be sacrosanct in the development of the AI system. Ethics demands that the matter of confidentiality of policyholder health information is given utmost consideration while designing AI software.

An important matter to be considered by a developer is, who bears the responsibility in the event of a lawsuit filed by a dissatisfied customer over the policy issued. The insurance company or the application developer or the machine itself? On the one hand and under the traditional concept, the insurance company, through its claim department, is responsible for the verification of healthcare information and management and issuance of an insurance policy; and, it is incumbent on them to exercise professional judgement and due care in determining the relevant policy. On the hand, the software developer may not be immune from liability for not anticipating possible scenario of a new case which was not inputted previously into the system. Hence, it is ethically right for contract agreement for the development of AI application to stipulate the owner of the ultimate liability in the event of litigation.

 AI presents opportunities across diverse industry and profession and, as seen above, the insurance industry is no exemption. However, certain ethical issues such as machine inflexibility, confidentiality over customer’s information and owners of responsibility for a failed system are sacrosanct and should be considered in designing an AI solution to automate health insurer decision to offer someone a new health

1.0       Habeeb, A. (2017). Artificial intelligence Ahmed Habeeb University of Mansoura, (September). https://doi.org/10.13140/RG.2.2.25350.88645

2.0       Hamid, S. (2016). The Opportunities and Risks of Artificial Intelligence in Medicine and Healthcare. The Babraham Institute, University of Cambridge, (Summer 2016), 1–4.

3.0       Matt Collier,Ricard Fu,Lucy Yin, P. C. (2017). Artificial Intelligence in Healthcare | Accenture, 1–8.

4.0       Mckinsey. (2017). Smart claims management with self-learning software Artificial intelligence in health insurance, (September). Retrieved from http://healthcare.mckinsey.com/sites/default/files/Artificial intelligence in Health Insurance.pdf

5.0       Rapaport, C. (2015). An Introduction to Health Insurance: What Should a Consumer Know?

6.0       Rolley, T. H. E. T., Ysteries, P. R. M., Kamm, B. F. M., York, N., Emens, E.,    Fanning, S., … Pozen, D. (2015). Book Review Law and Moral Dilemmas, 360(15), 659–699.

7.0  NHE Fact Sheet – Centers for Medicare & Medicaid Services. (2019). Retrieved from https://www.cms.gov/research-statistics-data-and-systems/statistics-trends-and-reports/nationalhealthexpenddata/nhe-fact-sheet.html

• What is confidentiality, integrity, and availability (CIA triad)? – Definition from WhatIs.com. (n.d.). Retrieved from https://whatis.techtarget.com/definition/Confidentiality-integrity-and-availability-CIA