Using deceased people’s personal data

Abstract

It is important to manage individuals’ personal data after their death to maintain their dignity or follow their wishes as much as possible. From this perspective, this report describes the real-world commercialization of immortal digital personalities, which gives eternal life to the deceased in a digital form. We identify the problems with the commercialization of deceased users’ images and personal data, which becomes postmortem entertainment. Considering these problems, we seek out the ideal form of deceased users’ personal data for commercialization. We conduct a social survey to understand how ordinary Japanese people feel about the various types of publicly available services that use personal data after death, such as social network service logs. By analyzing our survey results approximately 20% of respondents would allow the commercial use of their personal data, such as browsing their social network service logs, if they could receive compensation during their lifetime.

1 Introduction

The increasingly vast amount of data transmitted by Internet users through social network services (SNS) and email messages accumulates on platform providers’ servers over time. After platform users die, their data forms their digital heritage. However, users do not necessarily have a consensual understanding about how their digital heritage will be managed and by whom. Some users may wish their digital heritage to be deleted after their death, while others want their family members to manage their digital heritage. In addition, some users may wish their digital heritage to be read or viewed by many people and want to allow their digital heritage to be commercialized.

Internet platforms and service providers cannot only manage deceased users’ digital heritage but also profit from its commercialization. For example, Facebook (now Meta Inc.) provides a way for users to designate Digital Assets and a Digital Legacy Contact (Facebook 2022) and entrust the postmortem handling of the deceased users’ data to the nominated contact, who may possibly manage the resulting memorial account. If the number of users accessing the memorial account is large, there may be scope for generating advertising revenue. Several Internet platforms are also beginning to commercialize the use of digital heritage more directly (Öhman and Floridi 2017). However, some examples of parody have become ethically problematic, such as attaching animal ears to a photo of the deceased user.

Society is considering how to best manage deceased users’ digital heritage with respect for their wishes, especially by Internet platform business operators. However, we must also consider the possibility that people’s attitudes will change with changes in the development of technologies and social norms. Under these circumstances, it is meaningful to conduct a questionnaire survey using the general public regarding their ideal form of digital heritage in the future and analyze the results.

Section 2 describes the current status of managing digital heritage, while Sect. 3 introduces the idea of an “immortal digital personality” that uses deceased users’ personal data and how to realize this immortal digital personality. Section 4 summarizes the literature exploring the immortal digital personality and the related problems. Section 5 outlines the method we employed and the overview of the survey on 2,749 samples randomly selected by from around four million candidate respondents the Japanese survey company Intage Inc. holds, Sect. 6 presents the survey results and Sect. 7 analyzes the results. Section 8 summarizes our findings.

2 Background

2.1 Users’ intentions to keep their data after death

People are constantly accumulating a wide variety of digital data in modern society. After their death, these deceased users leave behind a large amount of personal data. These data are often referred to as the deceased users’ digital heritage. Following a 2012 survey of 2,046 people in the UK, Sofka et al. (2017) found that 20% of the survey participants had considered what should happen to their social media profiles after their death. Of these social media users, 43% wanted their profiles to be deleted, 20% preferred that their profiles should be kept active but with comments disabled, while 16% wanted to allow other users to comment on their posts. Finally, 20% of users were not sure about what should be done in the event of their death. In addition, 45% and 25% of the surveyed 18–24-year-old and over 55-year-old participants, respectively, wanted their personal data to remain available online.

Morse and Birnhack (2020) did a survey about digital personal data for 478 Israeli Internet users in 2017. 37% of respondents allowed their spouses access their e-mail and 33% of them allowed their SNS logs. 18% of them were aware of Facebook’s Legacy Contact or Google’s IAM (Google 2022) and 6% activated these online tools. 36% of respondents denied access to their accounts after their death by someone, 19% allowed access to some of the content and 45% allowed to access all contents.

After completing a similar survey, Orita (2019, 2021a, b) reported the following results for Facebook posts: an average of 50% of Facebook users in Japan, the US, and France wanted their information to be deleted, while 24% wanted their data to become a digital memorial and 27% wanted their data left alone. However, 68% of Japanese users wanted their data to be deleted. In a survey of Twitter users in Japan, the US, and France, 62% wanted their data to be deleted and 38% wanted their data left alone. However, 74% of Japanese users wanted their data to be deleted. In both cases, the number of requests for deletion was much higher in Japan than in other two countries.

Accordingly, scholars considered how deceased users’ posthumous digital heritage should be treated by surviving users. Chu (2015) proposed extending the scope of property rights to digital heritage by treating emails and SNS logs associated with personal accounts on Internet platforms in the same way as the deceased’s financial assets and other property. A significant portion of deceased users’ digital heritage can be accessed without inheritance policies by setting up Google’s Identity and Access Management (IMA) (Google 2022) or Facebook’s Digital Legacy Contact (Facebook 2022) for others to access the deceased users’ SNS logs. Apple (Apple 2022) also gives families the right to access their deceased family member’s Apple ID. Bickert (2017) described Facebook’s consideration of deceased users’ memorial accounts, while Winter (2010) proposed the granting of partial legal personality rights to the deceased users’ persona, which are artificial intelligence (AI) programs that manage their digital heritage according to the deceased users’ wishes. In addition, Kurihara (2020) classified digital artifacts and investigated the legal possibility of heirs’ access to the deceased users’ SNS logs based on the laws and regulations of Japan and other countries.

2.2 Status and classification of digital heritage

The following is a list of possible digital heritage formats, media, value, recipients, or users.

Physical storage locations: Cloud, personal computer, and smartphone data.

Format (fixed): Text, images, videos, and personal data.

Format (dynamic): Digital devices can be used as a system that responds to input, such as the big data from users’ long-term usage, which can be analyzed by AIs. The deceased users’ digital gaming skills are an example of these big data. Commercial services could provide their users with the opportunity to play against deceased expert gamers. In addition, deceased users’ digital art production skills could also be commercially relevant.

Media: Email, SNS logs, web pages, and software programs.

Direct monetary value: Financial institution account numbers, passwords, and crypto assets, etc.

Indirect monetary value: Items that could be converted into money through their use or sale, such as works of art (e.g., paintings, photographs, videos, or blueprints), self-portraits, information about the users’ dead bodies, gaming skills, and art-making skills.

Recipients and users: Bereaved families, relatives, and platform providers that the deceased used. However, platform providers are not limited to SNS, such as Google and Facebook, and also include commercial activities that specialize in managing the deceased users’ personal data, such as those described by Öhman and Floridi (2017) in their Appendix.

This study explores how deceased users’ digital heritage, i.e., their lifetime personal data, should be managed. Before their death, users may give their consent to allow their personal data to become informationally immortal if many people can use or interact with the deceased users’ data. In the following, we consider informationally immortal data as the deceased users’ immortal digital personalities, which the bereaved can use as a commercial memorial service. However, commercial use that goes beyond memorializing the deceased is also possible, whereas excessive use of digital heritage without any kind of regulation may lead to desecration of survivors’ memory of the deceased.

Privacy protection in the case of the living is concerned with avoiding inappropriate personal selection through data processing. After death, privacy protection becomes more about maintaining the deceased individuals’ dignity or following their wishes as much as possible. From this perspective, this study describes the real-world commercialization of immortal digital personalities, points out their problems, and explores ideal forms.

2.3 Legal issues

In addressing the legal issues related to digital heritage, Ulguim (2018) and Williams and Atkin (2015) referred to the concept of digital public mortuary archeology, which is the involvement of digital methods in the archeological content of the deceased. Following their discussion of this concept, an ongoing debate has explored the issue of digital heritage on Internet platforms within the theme of the digital archeology of death. According to Ulguim (2018), the ethical issues of publishing images of human remains online in archeology studies were debated intensely, such as at the 2013 Higher Education Academy event and by the European Association of Archeologists in 2015.

According to Harbinja (2020), the legislature regarding the deceased’s personal data, such as the 2018 UK Data Protection Act, defined personal data as “data relating to a living individual” and denies any rights to access deceased users’ personal data after their death. However, the European Union’s (EU) General Data Protection Regulation (GDPR) and Article 27, which is a new data privacy law, allows member states to introduce protections for deceased people’s data. As a result, the protection of deceased people’s data in the EU varies between countries.

In the United States, the 1986 Electronic Communications Privacy Act prohibits carriers from disclosing the contents of a deceased person’s communications without a court order. In addition, the Revised Uniform Fiduciary Access to Digital Assets Act (RUFADAA) considered that a deceased user’s last will and testament takes precedence in governing the disposition of their digital assets. This law allows the deceased individual’s personal representative to have default access to a “catalog” of the deceased users’ electronic communications and other digital assets that are not protected by federal privacy laws. In France, a similar law was passed, i.e., the 2016 Digital Republics Act. In the United States and France, the deceased’s wishes tend to be respected; however, the United Kingdom and Canada tend to respect the wishes of their heirs.

2.4 Direction of postmortem privacy

Harbinja (2020) considered that Conway and Grattan’s (2017) attempt to capture the treatment of digital estates as an extension of inheritance law did not sufficiently consider the deceased users’ privacy. To support her argument, Harbinja (2017) also used the concept of postmortem privacy, which she narrowly defined as the protection of personal data after death. If we regard the deceased users’ personal data as an informational body similarly to Floridi (2016), however, Öhman and Floridi (2017) showed the reality of commercial services’ use of proxy sites for the deceased. Figure 1 shows the relationship between personal data protection, narrowly defined postmortem privacy, and broadly defined postmortem privacy (Harbinja 2020).

Fig. 1

The relationship between basic privacy protection, narrow postmortem privacy protection, and broad postmortem privacy protection

Harbinja (2020) argued that legal solutions that promote the protection of privacy recognize the deceased users’ wishes as expressed through wills or technology should take precedence over the default application of traditional property and inheritance law. For better or worse, it is the technical aspects of digital heritage that facilitate this argument. Even if a deceased person’s wishes are preserved in written form, such as a will, it is only a small part of their total will, i.e., it relates to the distribution of their property after death.

Considering a surrogate to take over deceased individuals’ expanded will, including their life, purpose of living, and tastes Harbinja (2020) suggested that AI agents could be used in the future to represent deceased individuals’ will and their life before death, following Schafer (2010) and Buitelaar (2017). With the development of the immortal digital personality architectures and technologies described in Sect. 2, the challenge will shift to facilitating and constraining the use of these technologies under the control of the deceased users’ intentions (Öhman, 2018) or introducing a new law to cope with this problem (Harbinja 2021). In addition, Öhman and Watson (2021) discussed digital heritage, such as Instagram data, as having significant value for future generations. There is a need for a systematic curatorial framework for these data (Öhman and Watson 2021).

3 Immortal digital personality

3.1 Immortal digital personality structure

Among other scholars, Bell and Gray (2001) discussed the idea of achieving informational immortality through digital technologies, such as immortal digital personality. In this Sect. 1 will explain the technological implementations by Savin-Baden et al. (2017) and  Savin-Baden and Burden (2019).

Figure 2 shows the rough technical architecture for an immortal digital personality (Savin-Baden et al. 2017).

Fig. 2

Technical architecture for an immortal digital personality

Immortal digital personality creators collect individuals’ intentional and unintentional behaviors over a long period of time before their deaths, including tracking deceased individuals’ online history data, such as emails, voice mails, blog posts, SNS posts, global positioning system location data, bank account transactions. AI systems then use these data to learn the parameters for deceased individuals’ personality models. The AI system will then be able to make predictions based on the learned data about the deceased individuals’ personalities.

Considering the information flow directions in the upper left of Fig. 2, emails, SNS logs, and purchase and behavior histories have a unidirectional information flow, with a bidirectional information interaction with the outside world. There are also unidirectional systems; for example, Facebook memorial accounts retain the deceased users’ online digital existence, but the memorial page usually does not allow memorial posts from other users on these memorial accounts.

However, a natural language-processing system can be used for two-way information exchange with the outside world, such as contracting with other organizations, interacting with other users while they are alive, and communicating with other peoples’ immortal digital personalities (Fig. 2). The AI system would also be able to use the deceased users’ writing style to generate text.

After death, the deceased user’s personal appearance and AI system can be used to create a human-like digital avatar that can interact with the wider outside world, such as shown in the lower left side of Fig. 2.

From a philosophical perspective, it is important to note that the constructed immortal digital personality is not fixed. For example, the deceased’s diverse and context-dependent online history used in AI training represents a cross-section of the subject’s personality at any given time. In the present context, individuals’ long-term or universal personality is determined by the accumulation of the real-world personality’s temporal development and real-world contexts.

Immortal digital personalities are constructed with great care and effort using data and programming code. Their loss could be very damaging; therefore, many copies would be made to prevent this loss. By paying for these services, individuals can continue to live after death through their immortal digital personalities.

Immortal digital personalities do not necessarily have to be conscious because commercial digital heritage services can maintain the illusion of the immortal digital personalities as a model consistent with the way the deceased individuals behaved while they were alive.

3.2 Commercial survival of immortal digital personalities

Therefore, immortal digital personalities can live if the deceased’s heirs continue to pay an operating fee to the commercial company hosting the personalities. In addition, if the immortal digital personality can earn a profit from its operations, it can continue to stay alive. Nevertheless, countermeasures against the misuse or attack of immortal digital personalities must be considered. As an example of misuse, if a parent uses the immortal digital personality of their dead child to commit fraud, the United Kingdom’s 2006 Fraud Act will be applied.

If the immortal digital personality is implemented as software, it may be subject to various external attacks. Savin-Baden and Burden (2019) proposed the following measures to protect immortal digital personalities from external attacks:

• (1) Multiple copies and high-level redundancy.

• (2) Modular association structure to minimize damage.

• (3) Backup copies.

• (4) Monitoring and protection using high-level firewalls and antivirus software.

• (5) High-level user authentication for users with privileged access to immortal digital personalities.

• (6) High-level integrity checks to ensure that the code and data of immortal digital personalities are not corrupted.

Smart immortal digital personalities automatically implement these measures.

3.3 Examples of commercial services using immortal digital personalities

The following tools are some examples of immortal digital personalities that are commercially operated by companies: Eter9, LifeNaut, and Eternime.

3.3.1 Eter9

Eter9 (https://www.eter9.com/) is a social network entity that relies on AI as a central element. It is a virtual existence that interacts with the world by publishing, commenting on, and interacting with the deceased even when the target individual is absent or has already died. This virtual being acquires such aspects of the individual as their personality, intelligence, and skills through dialog with the individual during they are alive and continues to exist as a proxy for the individual after the individual’s death.

3.3.2 LifeNaut

LifeNaut (https://www.lifenaut.com/) provides a method for constructing a data structure called a Mindfile, which is used to collect and organize detailed data about individuals who are specifically profiled by answering a questionnaire with 486 questions. The specific profiles that LifeNaut measures include preferences, prudence, honesty, cooperativeness, and cohesiveness. In addition, LifeNaut also provides a function for learning and correcting users’ profile contents through conversations with avatars.

3.3.3 Eternime

Eternime (http://eterni.me/), started in 2014, was designed for individuals to train their own immortal digital personalities through daily interactions with the system before their deaths. The system then mines their personalities using personal data, such as Facebook posts, Fitbit recordings, Twitter posts, emails, photos, videos, and location data. Immortal digital personalities are formed using pattern matching and data mining algorithms and are retained even after death, which results in an immortal digital avatar. However, as of 2022, this service is suspended.

3.3.4 Virtual Barry

Virtual Barry (Savin-Baden and Burden 2019) is an example of the implementation of an immortal digital personality with a structure similar to Fig. 2. Virtual Barry is created through iterative interactions with the target individuals over a period of about 2 years. Specifically, information is collected manually through the following methods:

• (1) Face-to-face interviews and subsequent text-to-speech conversions.

• (2) Skype voice and text chat interviews.

• (3) Keeping the interview application running and answering externally inputted questions so that subjects can answer Virtual Barry’s questions whenever they want.

• (4) Completing a spreadsheet grid to ensure consistent data collection, on topics such as customers, projects, or employees.

• (5) Importing selected data from subjects’ cell phones, web browsers, directories, and calendars.

Considering their immortal digital personality system presented above, Savin-Baden and Burden (2019) discussed the mutual influence between the public space of constructing digital monuments and the realization of private memorials inherent in digital death. There are three modes in the memory and invocation of individual digital legacies: i.e., digital zombies, unidirectionality, and interactivity.

Digital zombies: The deceased users’ digital contents are activated without the deceased individuals’ permission, which results in the creation of “digital zombies” or “restless dead.”

Unidirectionality: Basic online accounts are managed as “digital legacies” that hold deceased individuals’ images after their death, while more advanced services provide full self-documentation, such as LifeNaut’s digital “one-way immortality.”

Interactivity: Digital technologies combine data and machine learning to create virtual avatars that provide immortal digital personalities that can be interacted with in both directions. These immortal digital personalities represent humans simply as “data patterns” or cybersouls.

According to Ulguim (2018), most online content includes little to no contextual data documentation, which poses a danger for the publishers and researchers who deal with this type of data. In addition, some three-dimensional (3D) images already had thousands of views and were available for reuse, modification, public download, and 3D printing.

If the semantic content of these collections of “poorly documented” contents are described poorly, the collection has little value as a tool for research, educational use, or public engagement, which can pose a significant threat to the potential for malicious or profitable use.

Kasket (2019) argued that technology designers and developers have a great deal of authority and responsibility regarding the moral decisions related to digital heritage management. With few ethical or legal guidelines in place, however, it is difficult to ascertain how digital heritage can be managed.

4 Current unfavorable problems and future directions for digital heritage management

4.1 Philosophical and economical background

Karppi (2013) gave the philosophical consideration about how immortal digital personalities are treated by the platform companies. The online locations where immortal digital personalities are realized is called deceased individuals’ nodes, which are operated by an online platform operator or a specialized company. Karppi (2013) described deceased individuals’ nodes as not only places to mourn the dead online, but also as nodes that are open and accessible to many people and other nodes.

First, Karppi (2013) introduced two types of politics: i.e., biopolitics and noopolitics. Biopolitics addresses the population’s economic, biological, and spiritual life, while noopolitics controls collective action and builds collective intelligence. If the deceased are positioned to unite the masses, then they become the means to achieve noopolitical goals. In this way, memorial accounts and pages accessible online should not be considered as merely places for viewers to gather and mourn, but as mechanisms that can also influence their actions, thoughts, and behaviors with the advancing information technologies, especially AI technologies used in AI system depicted in Fig. 2.

In this sense, the essence of deceased individuals’ nodes is not to guide viewers in remembering deceased individuals, but to help transform survivors’ memories of the deceased and use them to create new memories, which is the most important function of noopolitics.

Next, we consider deceased individuals’ memorial pages and accounts from the economical point of view. Their pages and accounts are not only there to represent the deceased individuals. When deceased users’ social networking logs, e-mails and so on become memorial accounts, they are “materialized.” Through this materialization, the memorial account gains economical use value and can be exchanged. The deceased cannot themselves participate in actions, share things, or contribute directly to the accumulation of user information. However, they can serve as navigation points for other viewers to participate in activities that originate from the deceased individuals’ actions before their death. Even if the navigation point is passive, it will be profitable for the platform companies. If the navigation point becomes an active AI agent as depicted by Fig. 2, the active navigation point would attract more people via interactive activities and finally make more money.

Previously, the deceased users seemed to be regarded useless for the SNS business model. However, scholars explore how deceased users can become part of the SNS platform or can be commercialized (Burden 2020). Along this line of usage of the deceased users’ accounts, Facebook, for example, wanted deceased users’ accounts to be memorialized (Bickert 2017) instead of deleted. In addition, if more people visit the deceased users’ memorial page, it will benefit Facebook’s advertising business. This is one of the expected courses from an economic standpoint.

4.2 Informational body and deplorable use

In addition to the biological body, Floridi (2016) insisted that individuals need to consider their informational body, which is constructed through integrating a variety of information sources that define human identities online, such as memories, biometric information, search histories, and social media data.

Information bodies should be supported by dignity and autonomy, and treated in the same way as postmortem bodies (Floridi 2016). In considering the informational body after death, Öhman and Floridi (2017) criticized the current situation where the use of the deceased users’ personal data has become too commercialized. First, the digital afterlife industry (DAI) refers to businesses that produce goods or services involving the online use of digital heritage to monetize deaths online. The incentive for DAI is to modify the online body of the deceased to increase its commercial value, which sometimes induces a violation of the principle of human dignity.

However, the reality is that most user-generated content of the deceased’s informational body after death is held by the Internet platform, which often acquires full rights to the deceased users’ information at the time of death because transferability of data on the clouds has not been established (Hopkins 2013). In addition, there may be an inherent tension between Internet users who wish to extend control over their personal data, such as their social networking logs and e-mails, after their death, and their bereaved families and friends who wish to use their personal data as mementos, for example (Morse and Birnhack 2020). Facebook had realized the tension between them and tried to alleviate it by introducing the scheme of account memorialization (McCallig 2014), and finally reached the concept of Legacy Contact.

Although Facebook’s Legacy Contact (Facebook 2022) and Google’s IAM (Google 2022) give users the opportunity while they are alive to decide what to do with their data on the cloud after death, however, the freedom of choices about that digital heritage are still not full-fledged. The following commercial services are currently being considered to deal with deceased users’ informational body, i.e., their digital heritage.

• Information management services and digital asset management are the basic infrastructure of various commercial services.

• Online memorial services deliver online messages or other digital communication content to designated recipients when a user dies.

• Postdeath messaging services provide an online space for individuals or groups to mourn and remember the deceased.

• Recreation services generate new content that replicates the deceased users’ social behavior based on AI technologies (e.g., Eter9).

This way of perceiving digital legacies are exactly what Floridi (2016) referred to informational body. And its use can be “deplorable”, so as to say, depending on one's position, and on one's short-term perspective.

When the deceased users’ personal data are used to generate commercial value, the deceased becomes an immortal digital personality who performs life activities (i.e., labor) to earn money from living Internet users. From this perspective, Öhman and Floridi (2017) discussed this problem in an interesting way. For example, company A presents an online version of the deceased that matches the behavior of the deceased while they were alive, while company B displays the most “consumable” version of the deceased, i.e., a modified version for online sales. Of these two examples, company B will experience greater market success. From a business perspective, it is impossible to prevent the deceased users’ personal data from becoming consumable. However, this is not unique to the recent management of deceased users’ personal digital heritage. For example, information about politicians, actors, and even charismatic religious leaders in some cases, such as their deeds and achievements during their lives, may continue to be used in ways that enhance their commercial value.

The role of technology in mediating the existence of the dead online is embedded in human practices, such as religious rituals to remember the deceased. Human practices are a matter of ethical decision-making behavior; however, when DAI intervenes, it is left to the DAI to make decisions about the deceased’s personal data. In this way, DAI should at least weigh their ethical constraints against economic rationality in managing the deceased users’ informational body. Economic rationality has both a short-term profit and a long-term aspect, i.e., the social reputation of the DAI. If, however, short-term profitability will be emphasized, it is expected to result in a situation similar to the desecration of the dead such as stated in the following two paragraphs.

When users go offline or die in Eter9, they live on as a Niner in the social network. Niners are AI agents that replicate the users’ behavior and manage their accounts when users go offline or die. Therefore, the AI agents make these deceased users’ digital heritage as consumable and profitable as possible for the living people to interact with these digital remains.

Öhman and Floridi (2017) shows another much more deplorable use. In 2016, Unilad.com published a funeral service using a photo of an open coffin. They even published a Snapchat filter to exchange one’s face with the deceased (Unilad 2016). In less than two days, the article garnered much attention, with more than 7500 shares. Almost 46,000 votes were cast on the question: “how fucked up is this on a scale of one to absolutely fucked?” This event was a huge revenue generator for Unilad.com. While the body of the deceased remained untouched, this example of a commercialized informational body part (a digital image) being altered and distributed was a gross violation of the dignity of the dead.

Considering the need to maintain the dignity of the dead, Öhman and Floridi (2017) suggested that DAIs should be subjected to the same discipline as archeological museums that display biological remains. While Öhman and Floridi’s (2017) idea of preserving the dignity of the dead is quite legitimate, the deplorable reality of the use of the deceased users’ personal data as described above is due in large part to human nature. Therefore, it is doubtful that simply calling for a ban will be effective. A line of demarcation must be identified that is ethical and acceptable to the majority.

Advances in AI technology will allow the AI systems shown in Fig. 2 in the Sect. 3 to perform more sophisticated and complex processing. For example, Burden (2020) predicted that AI technologies which will enable conversation via natural language will develop over the next decade. In addition, Eter9, LifeNaut, Eterni, and Virtual Barry, as well as the approximately 60 postmortem personal data services listed in the appendix of Öhman and Floridi (2017), will be able to further develop the functionality of their services. As a result, its use for could be “deplorable” and rampant. Hence, it is important to know what functions will be accepted by society in AI systems dealing with deceased information bodies in the future.

To this end, the first thing we have to know is whether ordinary general public tolerate what kinds of usage. This is why we conducted our survey described in the Sects. 5, 6 and 7.

5 Research questions and method

As described above, the commercialization of deceased users’ personal data is increasing in many countries. To the best of our knowledge, however, these services or any commercialization of deceased users’ personal data by Japanese companies do not exist in Japan. Nevertheless, Google’s Inactive Account Manager (IAM, an account deactivation management tool), Facebook’s Legacy Contact (a memorial account manager) (Facebook 2022), and Apple’s Digital Legacy (Apple 2022) were recently launched in Japan as services that include designating a person to handle deceased users’ data after their death. To determine the potential of commercial services using postmortem personal data in Japan in the near future, it is important to generally understand Japanese people’s honest feelings about how they want their personal data, such as their SNS logs, handled after their death.

To identify this potential, Orita and Yuasa (2020) surveyed Japanese university students about whether they wanted to allow their SNS accounts and those of their close friends and family, such as LINE, Twitter, and Instagram posts to remain active after their deaths. The results showed a desire to keep the data of those close to them but to erase their own data. A subsequent study by Orita (2021a, b) divided these SNS accounts into real-name and anonymous accounts, and surveyed 372 web monitors aged in their 20–60 s or older about whether they wanted to keep these two types of accounts active after the death of these users, and who they wanted to ask to delete them by clarifying the difference between the “main” account associated with real-life users and “sub” accounts associated with online relationships.

Based on these surveys and earlier studies, this study conducted an online questionnaire in an exploratory quantitative survey of the Japanese general public to clarify the following research questions:

RQ1: How do people want their personal data to be managed after their death?

RQ2: How do people allow their personal data, such as SNS message logs, to be used after their death?

RQ3: Do people prefer to receive any income from their digital heritage or not?

RQ4: What attributes, such as gender, age, or religion, make people’s decisions different?

RQ5: Does the frequency of Internet use make a difference in their decision?

6 Distribution of respondents’ attributes

In this section, we explain the survey method we used for the purposes described at the end of previous section. The online survey was conducted using the system and panels organized by INTAGE Inc., a Japanese research company. The survey was authenticated by JIS Y 20252 (an equivalent to ISO 290252, which is the international standard in a quality management system for organizations that conduct market, opinion, and social research). INTAGE Inc. has a set of potential respondents consisting of around four million Japanese people.

We requested INTAGE Inc. to use at least 2500 respondents from the four million potential respondents with as little bias as possible in terms of their age, gender, occupation, and place of residence. As a consequence, INTAGE Inc. used 2749 randomly selected respondents, as per our request. The survey was administered as a web-based, selective questionnaire, not open-ended ones, which respondents completed on their own.

Next, we present the results for respondent attributes distributions that could be used as explanatory variables in our analysis.

First, the gender distribution was 1376 men and 1373 women from a total of 2749 respondents. Table 1 shows the age distribution, where the average age is 50.2 years old.

Table 1 Age distribution of respondents

To analyze the relationship between the frequency of use of Internet services, we asked the survey participants about the frequency of their use of personal accounts for Internet services, such as emails, Twitter, LINE, Facebook, Instagram, and Google. This question was labeled Q-is. The respondents who answered that they do not use Internet services in this question were then instructed to answer the following questions on the assumption that they do use Internet services. Table 2 shows the distribution of their answers.

Table 2 Respondents’ frequency of internet service use

In all the tables shown in the remainder of this paper, the superscripts “a” and “b” and subscripts “a” and “b” indicate whether the number of persons in the cell is significantly more or less than the number of persons proportional to the distribution of persons by attributes on the horizontal axis shown on the upper first raw. In other words, significantly more or less is calculated within each raw. Superscripts “a” and “b” indicate “significantly more” at significance level of 1% or 5%, respectively. Subscripts “a” and “b” indicate “significantly less” at significance level of 1% or 5%, respectively. Intuitively, significance level of 1% cases shows stronger consequence than that of 5% cases. More precisely, 1% (or 5%) means the probability that the hypothesis “the number of people is larger (or smaller) than the number of people proportional to the distribution by attribute” does not hold, based on Z-score which is a variable standardized to have a standard deviation of 1 and a mean of 0.

Table 3 shows the relationship between the frequency of Internet service use and age distribution. In the age group of 18–29 years old, most of the respondents spend more than one hour on the Internet every day, and as the age of the respondents increases, the time spent on the Internet decreases except for respondents aged in their 70 s who usually have already retired and have enough time to spare.

Table 3 Cross-tabulation of frequency of internet service use and age (unit: persons)

In terms of the handling of personal data after death, we focused on the respondents’ attitudes toward religion. First, we asked the following questions, which we labeled Q-rel, about how the respondents dealt with religion in their daily lives.

Q-rel: The respondents were asked to choose from the following options to describe how they dealt with religion:

• (1) I believe in a particular religion.

• (2) I have no religious faith, but I follow the appropriate customs from different religions for weddings, funerals, and other ceremonies.

• (3) I have no religious beliefs, and my weddings, funerals, and ceremonies are conducted without religion.

• (4) I do not want to answer.

Table 4 shows the number of respondents who chose each option for this question.

Table 4 How respondents use religions

Next, the respondents were asked about their individual religious beliefs in the following question, labeled as Q-religion.

Q-religion: Please choose your religion from the following options: (1) Buddhism, (2) Shintoism,¹ (3) Christianity, (4) Islam, (5) other religions, (6) no belief in a specific religion, and (7) prefer not to answer.

Table 5 shows the number of respondents belonging to each religion.

Table 5 Respondents belonging to each religion

The results shown in Table 5 appear to contradict Table 4, where only 230 respondents adhered to a particular religion. This seems to be because many respondents answered very superficially to the Q-religion question. For example, many respondents answered “Buddhism” if they had been to a Buddhist funeral or if they had a Buddhist mortuary tablet. On the one hand, our cross-tabulation showed that 1,096 respondents had no religion and followed a religion only for weddings, funerals, and ceremonies, or had no religion for weddings, funerals, and ceremonies. On the other hand, no respondents said they followed a specific religion in response to Q-rel and no religion in Q-religion. These results confirmed our assumption. Therefore, it should be noted that Buddhism and probably other religions were selected superficially.

7 How personal data are managed after death

7.1 Respondents’ questions and answers

In this section, we describe the survey results exploring how the respondents would like their personal data to be managed and by whom after their death.

Q-del: For the real or anonymous personal accounts that you are currently using on Internet services, such as Twitter, LINE, Facebook, Instagram, or Google, you were asked to choose from the following two options for how you would like your personal information, including the accounts, emails, posts, and transmissions, to be handled after your death.

Automatic deletion by the service operator.

• (1) The account remains as it is.

• (2) The third row of Table 6 shows the results for this survey question.

Table 6 Cross-tabulation between Q-del and the frequency of use of internet services (unit: person; ratio [%] is for the total number of respondents for each internet use frequency as shown in the leftmost column.)

The cross-tabulation of these results with the respondents’ desired treatment of personal data after death and the frequency of use of Internet services are shown in rows 4–9 in Table 6.

Internet service providers automatically delete personal data.

In the case of real-name accounts, many respondents wanted to have their accounts automatically deleted because they do not want their family to discover their account and view it after their death.

In the case of anonymous accounts, those who use their accounts for more than an hour every day, say “heavy users”, are significantly more likely to prefer “leave as is” with a 1% significance level, while those who use their accounts for a short time every day, say non heavy users, are significantly more likely to prefer “automatic deletion” with a 1% significance level. Heavy users are likely to share their feelings and thoughts with many people via the Internet. At the same time, if they do not want their feelings and thoughts to be read by their family members, they likely set their screen name as pseudonymous nickname. Given these two factors, the author's conjecture is that they do not want to be automatically deleted simply to avoid being seen by their families. Non-heavy users, on the other hand, do not seem to have a strong desire to transmit their feelings and thoughts to many people through the Internet. This may explain the different attitudes of the two types of respondents to this question.

Q-real: We asked the respondents if there was anyone, they could recommend to manage their personal data, such as posts made to their current social network accounts on Twitter, Facebook, or Instagram after their death. Multiple answers were allowed, and Table 7 shows the results. The denominator for the percentage is the total number of respondents, e.g., 2749.

Table 7 People who can manage respondents’ internet services data after their death (unit: person)

On the one hand, the absolute difference in the number of respondents who can recommend someone to manage their data after death between the real-name and anonymous accounts is not large. However, at a 1% level of significance, significantly more respondents would ask for help if they had a real-name account than if they had an anonymous name. On the other hand, significantly more respondents indicated “no one to ask” and “do not want to ask” for their anonymous accounts than real-name accounts, with a 1% significance level. Comparing this result with the results in Table 6, respondents did not want to ask specific people to maintain their anonymous accounts, but they preferred service providers to leave them intact as memorial accounts so that they can be read by many people even after their death.

Q-wh: The respondents were asked to choose from the following options who they would like to read the data (e.g., posts, photos) left in their personal accounts and Internet services, such as SNS, after their death, in cases where the data are linked to their real-name account and where the data are not linked to them because the account is anonymous: (1) spouse, (2) children, (3) siblings, (4) SNS friends (e.g., followers), (5) others, (6) everyone (public), and (7) no one. Table 8 shows the results for this question.

Table 8 To Whom respondents willing to give their personal data read after their death (unit: person)

On the one hand, for both real-name and anonymous accounts, a certain percentage of respondents wanted to be read by their spouse, children, siblings, and SNS friends, but real-name accounts were significantly more common in the case of spouse, children, and siblings, with a 1% significance level. On the other hand, the number of respondents who did not want their anonymous accounts to be read by anyone was significantly higher, with a 1% significance level. The results of Table 8 showed the same trend as Table 7. These findings are a natural consequence from the fact that entrusting the management of their data after death assumes that it will be read by the assigned manager.

7.2 Summary and discussion

In this section, we discuss the results of RQ1: “How do people want their personal data to be managed after their death?”.

The results showed a similar tendency to the results of a survey conducted by Orita (2019, 2021a, b) in Japan, the US, and France to determine whether people want their Facebook, Twitter, and Instagram accounts to be kept active after their death. The results showed that when people accessed the Internet more frequently, they were more likely to want their Facebook kept active as a memorial account. These results are consistent with the results shown in Table 6.

8 Acceptable uses of personal data after death

8.1 Questions and answers from respondents

In this section, we present the survey results for the acceptable uses of personal data when it is made public after death. Multiple answers were allowed for acceptable uses.

The following eight use types were assumed and are directly connected to RQ2: “How do people allow their personal data, such as SNS message logs, to be used after their death?”.

The expressions in parentheses are used as shorthand for the items in the tables.

• (1) Allowing other users to browse their data (browsed by other users).

• (2) An AI system based on the users’ posting history, such as shown in Fig. 2, will answer questions from other users (AI dialog system).

• (3) Other users can add something to your posts (added text by others).

• (4) Users can add something to photos of themselves (added to self-portrait photo).

• (5) Other users can add something to a picture you took (added to photos taken by self).

• (6) Other users can modify a photo that you are in. For example, removing wrinkles and sharpening the image (modifying self-portrait photos).

• (7) Users can parody photos of themselves (parodying self-portrait photos).

• (8) Other users can modify a photo of your funeral (modifying funeral photos).

In the following three publication models related to RQ3: “Do people prefer to receive any income from their digital heritage or not?”, we asked the respondents whether each of the above eight use types was acceptable or not.

Q-noc: For the personal accounts of email, Twitter, LINE, Facebook, Instagram, and Google service users, these Internet service providers will release users’ personal data free of charge after their deaths and provide the above eight use types to the general public. However, no compensation is paid to individuals, such as the deceased users or their heirs.

Q-heir: Internet service providers obtain the deceased users’ personal data, disclose it to the public, and provide the above eight services to the general public. However, compensation is paid to the deceased users’ heirs.

Q-self: Internet service providers, including the information technology platforms listed in Q-noc, make public users’ personal data collected with the approval of service users while they were still alive and provide the above eight use types to the general public after death. However, since this approval was obtained before the users’ death, it is assumed that compensation is paid to users before their death.

Table 9 shows the results for the number of respondents using each of the above eight uses for Q-noc, Q-heir, and Q-self.

Table 9 Number of respondents accepting the eight use types for personal data in Q-noc, Q-heir, and Q-self (cross-tabulation by gender is also included; unit: person)

This table shows that at the 1% level of significance, Q-self had significantly more responses for browsed by other users (1) and responses by AI dialog system (2) than Q-noc and Q-heir.

Adding text by others (3) was tolerated by a fairly large number of respondents. The authors’ conjecture about this fact is that many respondents would not refuse someone to write condolences or other words on their social media accounts after their death.

The number of people who tolerate the usage of (3) to (8) is smaller than that for (1) and (2). This result may be due to the frequently observed situations that many respondents tend not to want others to write about their impressions of their deeds before their death and not to have their photos manipulated after their death.

The least accepted was parodying self-portraits (7). Even if they could earn money through allowing parodying, few respondents accepted because they wished to avoid having their dignity violated. In the order of no income, heirs’ income, and one’s own income, the number of respondents who allowed the use of (1) to (8) increased. In the case of Q-noc, there is no financial benefit to the respondents or their heirs, but in the case of Q-heir, heirs are paid. The significantly large number of respondents who allowed the use of their own personal data after death in Q-self indicates that many respondents valued the economic benefit to themselves, even if the use of their own personal data after death is a psychologically critical issue.

Considering gender differences, a significant majority of men tolerated all the uses in Q-noc, Q-heir, and Q-self.

When the correlations between Q-noc, Q-heir, and Q-self were calculated for the individual respondents who tolerated each of the eight use types, the correlations between Q-self and Q-noc were lower than those between Q-noc and Q-heir and between Q-heir and Q-self. Table 10 shows the results.

Table 10 Correlation coefficients between Q-noc, Q-heir, and Q-self for nine different uses of personal data (correlation coefficients are significant at the 1% level)

8.2 Analysis

The eight use types for personal data asked about in Q-noc, Q-heir, and Q-self provided survey results that predict the future of the actual use of the deceased’s personal data. It is assumed that people decide how to use their own SNS logs and other personal data after their death while they are still alive. Therefore, by examining the cross-tabulation between the answers to these three questionaries, which reflect their own intentions, we can predict the future image of acceptable postmortem uses of personal data. Therefore, we will discuss the results of the survey by cross tabulating the number of people who accept each of these eight uses to the questions with the results of the other questions.

To reduce the redundancy, the following analysis is based mainly on the cross-tabulation table with the results of Q-self, which is greatly influenced by the intentions of the respondents and involves the sensitive issue of compensation during their lifetime. Table 11 gives the answer for the RQ4: “What attributes, such as gender, age, or religion, make people’s decisions different?”.

Table 11 Acceptability of postdeath use of personal data by age group (unit: Person)

The number of respondents aged 18–29 years who allowed all eight use types was significantly higher, which suggests that they were less resistant to these uses. The number of respondents who allowed browsing among the 417 respondents in this age group was 81 in Q-self (19%), which is the largest among Q-noc, Q-heir, and Q-self. In addition, the number of respondents who allowed AI dialog system responses was 44 in Q-self (11%), which was the largest among Q-noc, Q-heir, and Q-self.

Parody was allowed by 13 respondents in the largest Q-self group (2–3% of the total), which indicated their high-level resistance to commercialization. The number of respondents aged in their 40 s was slightly lower than for respondents aged in their 30 s, while the number of respondents aged in their 50 s and above was very low.

Table 12 shows a cross-tabulation between the frequency of Internet use and the acceptability of postmortem use of personal data, which answers RQ5: “Does the frequency of Internet use make a difference in their decision?”.

Table 12 Acceptability of postdeath use of personal data in relation to frequency of internet use (unit: person; the denominator of the ratio expressed as a percentage is the number of acceptable users for each usage method in Q-self (listed in the leftmost column)

The so-called heavy users who use the Internet for more than one hour every day were significantly more tolerant than those who use the Internet less frequently. They also tolerated (3)–(5) use types. Although the results are not shown in this table, photo modification (6), photo parodying (7), and funeral photo processing (8) were also tolerated by significantly more respondents than the total number of respondents in each use frequency. Nevertheless, they were still in the minority, with less than 20% of respondents in Q-self browsing and 8% in AI use.

Compared with this heavy user group, those who used the Internet only for a short time every day were less tolerant of use methods (3)–(8). Light users and those who do not use the Internet at all had little tolerance for all use types.

Table 13 shows the results for examining whether the eight use types were acceptable for respondents who wanted automatic deletion and those who wanted to keep the account as is for each of the real-name and anonymous accounts.

Table 13 Acceptability of postdeath use of personal data for automatic deletion and leave as is categories

Surprisingly, a large number of respondents who wanted automatic deletion also allowed various uses of their data. It seems contradictory to allow postmortem use when automatic deletion is desired; however, this can only be interpreted as the result of thinking about the eight use types independently from their answers to the question: “want automatic deletion/leave as is?” In addition, the number of respondents who wanted automatic deletion or to leave it as is in Table 13 were proportional to the number of respondents who tolerated the various postmortem services. Statistically speaking, the ratio of the number of respondents who allowed each usage method to the total number of respondents who wanted their accounts to be deleted automatically was significantly close to the ratio of the number of respondents who allowed each usage method to the total number of respondents who wanted their accounts to be left as is. This suggests that our above assumption is correct.

From a statistical perspective, we also identified the following results. In terms of the ratio of the number of respondents who want their accounts to be deleted automatically and those respondents who want their accounts left as is, for anonymous accounts, the number of respondents who allowed the use of browsing, AI dialog system responses, writing comments, adding to photos, and adding to self-portraits was significantly higher among those respondents who wanted to leave their accounts as they are. The trend in the case of real-name accounts was similar to that for anonymous accounts; however, somewhat fewer respondents wanted to have their accounts retained but were also significantly more likely to allow all use types (Table 13).

We asked the respondents about whether they would use deceased people’s personal data to make a profit if the data were inherited or where the deceased individuals’ intentions regarding the management of their personal data were not clearly expressed, such as erasure. Three hundred eighty-one (13.9%) respondents said yes, while 2,368 (86.1%) said no. Eighteen out of 78 respondents with a professional occupation agreed that it was permissible to make profits was significantly higher at a 5% significance level and the largest absolute number was 109 out of 706 company employees.

Table 14 shows the cross-tabulation between the answers to the above question and the eight use types of personal data. The respondents not counted in Table 14 include 246 who would make a profit and 2,087 who would not. These respondents answered that there was no usage method that comes to mind among the eight use types. This table shows a part of the answers to RQ3: “Do people prefer to receive any income from their digital heritage or not?”.

Table 14 Number of respondents who want to profit from deceased users’ data

The number of respondents who allow all eight use types at the significance level of 1% or less was higher than the mean of 13.9%. The number of respondents people who do not allow the eight use types was lower than the mean of 86.1%. In other words, those who would profit from the deceased users’ personal data were more likely to accept the eight use types.

The results for respondents’ answers to the religious attitude question, Q-rel, and the acceptability of the eight use types are interesting. Therefore, the results of Q-rel and Q-noc, Q-heir and Q-self were cross-tabulated in Table 15.

Table 15 Acceptability of postdeath use of personal data in relation to how one treats religion (unit: persons)

The respondents who followed a religion only for weddings, funerals, and other ceremonial occasions were significantly more likely to accept the use of browsing and the use of the AI dialog system in response to the question in Q-self.

On the one hand, the respondents who followed a particular religion were significantly more likely to accept the following use types (3)–(5). On the other hand, there is no usage that was tolerated by a significantly larger number of respondents with no religion. We consider that the irreligious believe that they will disappear completely following their death. In other words, they are likely to be indifferent to what happens after their death. However, it is somewhat surprising that not many of these respondents were willing to allow their personal data to be used after death, which suggests that there is not a strongly positive psychological correlation between religious attitudes and the use of personal data after death.

In addition, we were interested in the relationship between the results for the respondents answers to the Q-religion that they actually believed and the answers to Q-noc, Q-heir, and Q-self. Hence, a cross-tabulation of these results are shown in Table 16.

Table 16 Acceptability of postdeath use of personal data for religious adherents (unit: persons)

For respondents with a religion, the acceptable use types did not vary by religion in any distinctive way. Surprisingly, the number of acceptable uses was not significantly higher for both respondents with no religion and respondents who did not answer the survey question, despite their large numbers. In summary, the number of acceptable uses did not change significantly depending on the respondents’ religion or beliefs. Tables 15 and 16 show parts of their answers to RQ4: “What attributes, such as gender, age, or religion, make people’s decisions different?”.

8.3 Summary of discussion

This section summarizes and discusses the findings and analysis presented in Sects. 8.1 and 8.2 and presents the responses to RQ2–5 presented in Sect. 5.

For RQ2: “How do people allow their personal data, such as SNS message logs, to be used after their death?”, less than 10% of respondents would allow their personal data, such as SNS logs, to be made public and readable after their death. In addition, 4% to 5% of respondents would tolerate interacting with an AI dialog system using their personal data. Only about 2% of respondents would tolerate any other public use of their own data. In other words, respondents have little desire for postmortem use of their personal data. However, it is not clear whether the respondents considered this as a negative reaction regarding postmortem use of their personal data, or whether they reacted negatively because they could not imagine a specific use for their data. In Japan, few services commercially use postmortem personal data; therefore, the respondents may understand the literal expression, but could not imagine specific uses for their personal data after death.

Considering RQ3: “Do people prefer to receive any income from their digital heritage or not?”, the percentage of respondents who approved of the public use of their personal information was still small (Table 9). In addition, a significantly larger number of respondents would allow their personal data to be disclosed publicly and used for such functions as browsing and interaction with AI dialog systems, if they themselves could be compensated compared with whether they could not be compensated (Table 9).

Considering RQ4: “What attributes, such as gender, age, or religion, make people’s decisions different?” and RQ5: “Does the frequency of Internet use make a difference in their decision?”, younger respondents were more likely to approve of posthumous disclosure. In addition, respondents who were more active on the Internet were more likely to approve of posthumous disclosure (Tables 11 and 12). These findings show that younger respondents were more active Internet users. Furthermore, younger respondents were more likely to allow the use of their personal data, which suggests that the percentage of people who allow the use of their personal data is likely to increase in the future. However, if the level of acceptance of postmortem use of personal data is greatly influenced by age and tends to decrease with increasing age, we cannot consider that the number of people who accept the postmortem use of their personal data will increase in the future. Unfortunately, it is impossible to obtain a clear answer by comparing these two possibilities because the current elderly population was already quite old by the time the Internet became widely used.

Considering RQ4: “What attributes, such as gender, age, or religion, make people’s decisions different?”, similarly to gender, male respondents were slightly more tolerant to the use of their personal data after death (Table 9). Considering religion, more pragmatic respondents, such as those who believed in religion as a formality for weddings and funerals were more tolerant to the use of their personal data after death (Table 15). Considering specific religions, Buddhists appeared to be the most tolerant (Table 16), but as already mentioned, those respondents who followed religions only for weddings and funerals were more likely to say that they were Buddhists. Thus, we did not find a tendency for religious beliefs to be closely related to tolerance. Instead, there was a trend toward lower levels of tolerance among respondents with no religion. Unfortunately, the survey results did not provide data that would allow us to interpret this trend.

9 Conclusions

In the first half of our study, we summarized the methods for dealing with postmortem personal data, including the undesirable ways in which postmortem personal data could be used for profit, and the knowledge about ways of using postmortem personal data. In the second half of our study, we empirically analyzed the acceptable ways for personal data to be managed after death. On the one hand, the results suggested a strong impact from financial returns. On the other hand, the use of the personal data after death, such as interacting with an AI dialog system built using personal information or adding something to a photo, was much less acceptable.

The limitations of our study are that respondents may not have a sufficiently concrete understanding of the use of their personal data after death and may have answered without being able to imagine it. This finding was limited to Japan, where respondents were reluctant to keep their personal data available after their death. Future studies could conduct quantitative and qualitative surveys to review the use of practical services to manage personal data after death to explore the use of data and monetary returns by considering individuals’ perspectives of life and death, in addition to their willingness to “resurrect the deceased” virtually. Furthermore, expanding the scope of the survey to other countries and regions would reveal issues that should be confronted by Internet services globally.