Big data from the perspective of business ethics

25. Mai 2022

With data increasingly available and analysable thanks to big data technology, private companies are seeking ways to harness it for commercial purposes. This development is perceived with both hopes and fears. In this commentary, the topic is addressed from the perspective of business ethics. Based on use cases, recommendations for action are formulated for firms that use or want to use big data.

Big data has been likened to a currency of the future (Eggers et al. 2013). Digitalisation has allowed not only for the mass creation, storage, and analysis of large amounts of heterogeneous data but also the commoditisation of this data. The use of digital technologies generates data on processes and areas of life that were not observable in the pre-digital age—for example, one can compare rummaging in a brick-and-mortar bookstore with searching for books on Amazon, where each online movement and click leaves a digital data trail. In general, every interaction with digital technologies creates a digital trace, and those traces can be combined to paint an intimate picture of everyday lives, habits, and interests. This is because digitalisation not only makes it easier to collect and store data but also because it has become increasingly inexpensive and convenient to analyse huge data sets, especially with the help of powerful (self-)learning algorithms. Thus, data on production and consumption processes have become commercially exploitable in a way that was almost unthinkable just a few years ago (Christen et al. 2019a).

Such drastic changes arouse both hopes and fears. Some emphasise the huge economic potential of big data, calling it the “oil of the 21st century,” that is, an enormous resource for innovation, progress, and wealth creation. Others consider big data to be a fundamental threat to freedom and privacy—a demonic instrument of an Orwellian surveillance regime (Helbing 2015). Neither position is nuanced enough, but they reflect the tensions surrounding the ethical, legal, and social issues (ELSI) of big data. Several aspects of this ongoing discussion have already been covered in detail in this white paper. These include the ethical challenges of using big data applications in certain sectors, namely healthcare (see Ienca’s and Elger’s main articles); values threatened by big data, such as privacy, autonomy, and transparency (see Christen’s main article); and fairness and non-discrimination in big data use (see Loi’s main article). In this commentary, I will expand on these issues. For this purpose, the topic of big data will be addressed from the perspective of business ethics. Recommendations for action, based on some concrete use cases, will be formulated for private sector companies that use or want to use big data technology.

With data increasingly available, private companies are seeking ways to harness it for commercial purposes. Use cases for the commercial exploitation of big data applications include, but are not limited to, (1) avoiding payment defaults, (2) improving risk management, (3) tailoring offer conditions, (4) enhancing the efficiency of advertising campaigns, and (5) creating innovations and opening up new revenue streams (Hauser et al. 2017).

First, to forecast the likelihood that a customer will pay on time and not default, companies have traditionally relied on the credit ratings of their customers. Big data-based scoring models are being used more and more frequently for this purpose. Algorithms analyse a plethora of (often unrelated) data points such as behaviour and interactions on social media platforms, browsing behaviour on search engines and websites, and the technical specifications of devices that are used to access the Internet (e.g., desktop computer, laptop, tablet, or smartphone) to determine a score. In online shopping, this score can be used to determine whether a particular customer can pay by invoice or must prepay. In banking transactions, the score can influence the terms on which a loan is granted.

Second, the business models of some companies (e.g., those in the insurance industry) are reliant on successful risk management (Christen et al. 2019b). Big data allows for a more accurate, case-by-case assessment of risks such as the probability and degree of damage. For example, telematic solutions used by the car insurance industry collect information on how a policyholder drives to calculate the likelihood of a collision. Policyholders of health insurance companies grant their insurers access to fitness data collected by wearables and smartphones. These data can be used by insurance companies to predict their clients’ health risks more accurately and offer tailored insurance products.

Third, dynamic pricing is a standard practice in several industries, used, for example, by airlines, e-merchants, and gas stations. Such industry players dynamically change prices based on data points such as demand, availability, time of day, and the behaviour of competitors. Big data allows companies to go beyond these traditional approaches and to predict the “ideal price” of a good or service for each customer by using both personal and technical data on each individual. For example, an online tour operator displayed higher prices to Apple users, because they tended to be less price-sensitive when booking (Mattioli 2012).

Fourth, advertising is effective when it reaches customers who are interested in a given product, but traditional advertising campaigns tend to use the “shotgun approach” (e.g., billboard advertising). Campaign accuracy can be improved with the help of big data analysis. In e-commerce, it is common practice to show adverts to (potential) customers based on their search or purchase history as well as their geographic location and other variables. With the advent of digital billboards, individualised ads that were previously confined to personal devices can in principle also be displayed on billboards to target people in the vicinity. Additionally, the advertising chosen can be tailored to the emotional state of the customer by an artificial intelligence-based application that analyses visual and auditory information.

Finally, companies can use big data to generate innovation. For instance, data analysis can identify emerging trends and prompt the development of new products (e.g., in manufacturing or streaming services). Similarly, data from voice recognition software can be utilised to improve voice-control technology or related services such as translation. Big data also enables advances in infrastructure and mobility planning that can help to minimise traffic congestion, for example. Moreover, thanks to big data, new business models are emerging, and companies can tap into new sources of revenue. For instance, companies can make their data available to other businesses and thus complement their product portfolio.

These use cases show the wide range of big data applications in the private sector and illustrate how big data is transforming the way that companies from different industries do business. These applications may produce added value for companies and their customers but also entail significant ethical risks. The ethical discourse has identified eight values that are affected by big data applications: (1) privacy protection, (2) equality and non-discrimination, (3) informational self-determination, (4) control of one’s (digital) identity, (5) transparency, (6) solidarity, (7) contextual integrity, and (8) property and copyright (Hauser et al. 2017). Each of the use cases described above touches on these eight values to some extent. Companies are advised to manage these risks carefully as customers and other stakeholders are likely to be critical of the violation of these values. Additionally, companies must demonstrate responsible data handling in their big data applications to obtain or retain their license to operate. Therefore, they should consider the following points when dealing with big data.

Take the “ethics case” into account

When evaluating (novel) big data applications, companies should not only focus on the technical feasibility and business case of the application but also its ethical implications; they should embed ethical considerations into the design process from the beginning. By systematically including the ethics case in the development process, companies can understand where potential violations of ethical values and conflicts may arise and resolve them proactively. Looking at the use cases described above, the following ethical deliberations might be considered.

To ensure privacy, data should only be collected for a specific purpose, in accordance with the principle of purpose limitation. Furthermore, according to the principles of data avoidance and data economy, only data relevant for achieving this purpose and not exceeding it should be collected. Big data applications often collide with these principles, as it is precisely through the combination of data sets from different sources that new insights can be gained. This exposes the risk that personal data are being used beyond the scope of their intended purpose, thereby violating the privacy of consumers, which can lead to a loss of legitimacy for a company.

The merger of datasets from multiple sources raises further questions around the control of the individual’s digital identity and right to self-determination of personal data. The phenomenon of digital identities arises from the ability of big data applications to aggregate and analyse multidimensional data. Problems arise when an individual is unaware that his/her data are being used in this way or has no means to change the categories within which his/her digital identity has been placed. Further, an individual’s identity is fluid over time. His/her preferences and interests change, rendering much personal data time and context dependent. Individuals also create silos for their online presence (e.g., a profile created on a professional networking platform versus a profile on a dating app). If this data is merged, the blurring of the boundaries between the different contexts may lead to inaccurate digital identities. In addition, these digital profiles open up the possibility for the violation of informational self-determination. For instance, data might be used for targeted advertising that aims to manipulate the individual (e.g., emotional advertising).

Making advertising more effective might constitute a legitimate business reason to analyse large data sets on customers. In this case, the benefits for the customers (e.g., seeing ads with relevant information) must be balanced against the drawbacks (e.g., digital “surveillance” by the company). From an ethical perspective, an intrusion into the customer’s privacy seems justified, if he/she is informed of the magnitude of data collection, has consented to it, and has a realistic alternative to consenting. However, it is problematic if customers who insist on their right to informational self-determination are denied access to certain services because they refuse access to personal data or if they are charged substantially more, resulting in de-facto discrimination.

A further challenge relating to big data analysis is the potential violation of the solidarity principle that often guides the insurance industry. When aggregating data for risk management purposes, the principle of solidarity is open to violation if there is the possibility of extending the cost-by-cause principle. If a given risk (e.g., of lung cancer) can be attributed to a behaviour that is chosen freely (e.g., smoking), then the decision to engage in the risky behaviour may become a reason for denying solidarity. Companies could even require certain (positive) behaviours from individuals to mitigate risks. This form of influence might seem economically attractive (e.g., for insurance companies who wish to reduce risks). However, it is in direct conflict with the right to self-determination and free will (Loi et al. 2021).

These deliberations show that it is of central importance for companies to consider the ethics case at an early stage in the development of big data applications. This by no means leads in general to the inhibition of big data applications. Rather, it makes it possible to develop more sustainable big data applications that would have a long-term “license to operate” from the stakeholders. Such an approach is therefore recommendable from an economic perspective.

Consider the customer’s point of view

A simple test criterion for assessing the “license to operate” of a big data application is the following: Would the customer consent to sharing his/her data if he/she knew what would be done with the data? If full transparency is treated as a standard baseline for development, companies must consider whether customers would be comfortable with the product or service that is to be developed utilising their data. This highlights the importance of maintaining transparency and clarity on the collection and use of personal data. Depending on the service in question, opt-in data solutions and the provision of acceptable alternatives may be attractive strategies.

This raises the question of the actual level of knowledge of the customers with respect to big data. The statements about a company’s analysis and usage of an individual’s data are often enshrouded in lengthy terms and conditions and legal jargon, not typically understandable by the average person. The algorithms used for big data analysis are usually proprietary to a company, meaning that the data used by the algorithms cannot be validated for quality, trustworthiness, or completeness. As a result, tensions emerge between a company’s rights to preserve its algorithms (i.e., the intellectual property connected with them) and customers’ rights to transparency.

Another consideration is what benefits customers would consider adequate in exchange for their data. Each interaction with online services creates a digital trace that generates and/or discloses some form of personal data. Typically, companies use the personal data collected from or about their customers for business purposes, with this data additionally serving as a basis for new revenue sources. This raises the question of the ownership and value of the data. Are the data generated through digital interactions considered creations in the sense of copyright law, or is this only the case when the data are analysed? The answers to these questions determine the attractiveness of the business case for the use of data. For example, it could be required to share revenues earned from data use with the customer. Anticipating the consumer reaction to the specific use of data can help avoid potential violations of property and ownership expectations.

Create transparency and freedom to choose

Big data applications cannot be widely and successfully implemented unless they are trusted and accepted by consumers and other stakeholders. This requires that companies provide transparent and comprehensible information about how they collect and use data.

With this in mind, the development of a big data application must involve a conscious, case-by-case ethical evaluation, considering whether there is a (potential) infringement of ethical values and, if so, whether the infringement can be justified. Whilst this may become burdensome for single (small and medium-sized) companies, it is recommended that collaborations for ethical evaluation be fostered and industry standards created to support this process (see, e.g., Christen et al. 2019b, 2020). At a meta level, it would be advisable to work towards standardisation in the terms and conditions relating to the collection, analysis, and use of big data, not only to guide company business practices but also to foster customers’ capacity to provide informed consent. The goal should be to empower customers as well as society to better comprehend the scope and scale of data collection, analysis, and usage in the age of big data.

Acknowledgement:

This commentary is based on earlier works by the author and his colleagues. Language assistance by Eleanor Jehan is gratefully acknowledged.

Recommendations

Big data applications can only be successfully implemented if they are trusted and accepted by wide parts of society. Policy makers should set the framework under which data can be collected, analysed, and used. They should not only insist on the creation of (self-regulatory) standards that balance the interests of firms and consumers, but also empower customers to make informed decisions.

References

Christen, M., H. Blumer, C. Hauser, and M. Huppenbauer. 2019a. The ethics of big data applications in the consumer sector. In M. Braschler, T. Stadelmann, & K. Stockinger (eds.), Applied Data Science: Lessons learned for the data-driven business : 161–180. Cham, Switzerland: Springer.

Christen, M., F. Thouvenin, C. Hauser, et al. 2019b. Big Data Ethics Recommendations for the Insurance Industry. Zürich: NFP 75.

Christen, M., C. Heitz, T. Kleiber, and M. Loi. 2020. Code of Ethics for Data-Based Value Creation. Thun: Swiss Alliance for Data-Intensive Services.

Eggers, W., R. Hamill, and A. Ali. 2013. Data as the new currency. Deloitte Insights.

Hauser, C., H. Blumer, M. Christen, L. Hilty, M. Huppenbauer, and T. Kaiser. 2017. Ethische Herausforderungen für Unternehmen im Umgang mit Big Data. Zürich: SATW.

Helbing, D. 2015. Thinking ahead-essays on big data, digital revolution, and participatory market society. Cham: Springer.

Loi, M., C. Hauser, and M. Christen. 2021. Highway to (digital) surveillance: When are clients coerced to share their data with insurers?. Journal of Business Ethics, 175, 7–19. doi:10.1007/s10551-020-04668-1.