21st Century Legal Issues
Professor David Friedman
In a virtual world, people will want to create virtual institutions—virtual partnerships, corporations and marriages—so that they can share virtual (and possibly real) assets. This raises a whole slew of issues. Among them are how to implement these partnerships, virtual property rights, eligible marriage partners, virtual children, limits to reproduction, etc. This paper mainly addresses virtual marriages, virtual property transfers after virtual or actual death, and family relationships in Virtual Reality (VR), with an emphasis on issues raised by the creation and raising of virtual children.
Before addressing the issues raised by virtual relationships, it is necessary to consider why anyone would want to have them in the first place. After all, some might argue that it is ridiculous to shift our relationships out of real space and into virtual space. Many people might become so isolated from reality that our society as a whole will suffer. One might imagine a world where everyone looks perfect in VR and will be embarrassed to be seen in public because they cannot compare to their virtual selves. If it is impossible to get any physical exercise in VR, then it might become normal for people who spend most of their time in VR to become immobile due to lack of physical strength and morbid obesity. Emergency workers may have to become accustomed to having to remove walls when removing the bodies of dead VR addicts. In spite of advances in medicine, we may see a surge of people ignoring their health and dying earlier than necessary, simply because they are too immersed in their virtual lives to take care of their health in the real world. We could also see a new form of anorexia, where people are too occupied with VR to remember to eat.
On the other hand, advanced VR systems might include taste and smell interfaces while allowing intravenous feeding of advanced foods that extend life, coupled with nanomachines that care for the body and machines that exercise the limbs to keep the body fit. This would eliminate the physical problems that might arise from living in a virtual world (and probably substantially reduce deaths from accidental injury.) But it might not address the social problems that might arise from people never actually physically interacting with each other in real situations.
However, one could imagine a variety of scenarios where VR is a useful construct. For instance, some people may only exist in VR (disembodied artificial intelligences (AIs) or humans who avoid death by transferring their consciousnesses into virtual worlds) or have no ability to function in the real world (aware but unable to interact in real world.) Some people may simply prefer to live in VR, maybe because they are ugly or misshapen in real world—or maybe because they have handicaps which are inconvenient in the real world, but non existent in VR. For instance, in a world where VR has a direct brain interface, deafness, blindness, missing limbs, and other physical defects may not exist at all.
Alternately, if our society becomes highly risk adverse, then virtual reality might become a preference. In a time of super-plagues or nano-weapons, people may seal themselves in protected environments and only interact virtually. Other grim future scenarios include ecological disasters, where no one can leave their shelters, or famines where the only available food tastes too bad to actually eat—but where you can have anything you want to eat in VR while having machines feed you goop that you never actually have to taste.
What if our society becomes dispersed throughout the solar system (or multiple solar systems)? VR could be a solution to alleviate the isolation that might come from long periods of isolation in remote outposts or long space voyages—especially if we somehow figure out how to avoid the long latency times caused by ultra long distance communication.
Other uses of VR might be for convicted criminals, who are too dangerous to allow to interact with people in reality, but harmless in VR, or aliens who interact at different time rates or who are just too scary-looking to be dealt with in their true forms.
Another possible use of VR is for people who have been cryonically preserved, assuming that brain function remains at some much slower rate. A VR world populated entirely by cryonically-slowed individuals would offer an opportunity for people to have virtual lives while waiting to be cured and restored to actual reality. Another possibility is that cryonically preserved people might download their brains into VR while waiting for a cure in the real world, with the intention of transferring the virtual experiences back to the real body when their real body is ready. This scenario also works for long starship voyages where the virtual world is the starship’s computer and the crew’s real bodies are in suspended animation during the voyage.
Having established that there are possible scenarios where virtual reality might be a useful substitute for actual reality, we can move onto the issues raised by life in a virtual world. For instance—do we want to transfer real-world institutions into VR? Given that, in many of the scenarios discussed in the last section, VR has become a substitute for actual reality, it seems that we would want to do just that. But it is obvious that many of the limitations of the real world will simply not apply in VR. Some examples: Many “people” in VR may be immortal; Real Property as a concept will have a completely different meaning; Virtual “children” may come into existence by non-biological means; Real world governmental institutions may be inappropriate or completely ineffectual in VR; “Death” as a concept will have to be redefined; Since some portions of Virtual worlds may experience time at different rates, concepts of time will have to be broadened; The issue of slavery will have to be revisited; and new problems may arise from population control pressures.
The focus of this discussion is on virtual marriage and families—including possible mechanisms for administrating virtual estates, raising children and various forms of alternate marriage that may work in a virtual society.
Virtual Marriage is probably one institution that will be desirable to implement in any deep virtual reality society, especially when considering the transfer of wealth after death, sharing of virtual properties during life, and to aid in the raising of virtual children. It might not be too useful in the raising of real children, however, since the process is probably better executed in the real world in any but the most extreme environmental conditions, in cases of great distance, or in cases of a severely physically impaired family members. If there is a usable real world and none of the participants are anonymous or non-corporeal, standard models of marriage will probably be sufficient. However, as soon as any one of the family members is unable to function outside of VR, then there will have to be a way to care for those who must remain in VR for their entire existence.
Virtual Marriage is also a potentially useful concept in a world of strong privacy. Say you have built up a large bank account that is completely untraceable to you. What happens when you die? It is pretty obvious that probate won’t help you pass your money onto your descendents, so an alternative means of passing money after death is desirable.
Implementing the actual mechanism for marriage is probably the smallest part of the problem—it will probably be no more complicated than getting a marriage license is in the real world. Alternate marriage types, such as line marriages, domestic partnerships and group marriages may be implemented as well, depending on the mores and values of the virtual environment.
Complications arise when dealing with issues of identity and life status, however. In a virtual world, it becomes very difficult to verify that the person you have married is actually one person, or is the sex you think he his, whether he is married in some other world, or even if he is human. If VR is tied to the real world by means of public keys, for instance, there is no guarantee that the virtual person you are interacting with isn’t actually an impostor who has stolen your spouse’s identity (giving a whole new spin on the meaning of “identity theft”.) In order for virtual societies to function properly, then, it will be necessary to implement a stronger system for verifying identity.
One can conceive of various ways to identify a person in VR. If the person is a flesh and blood human, then some combination of DNA, voiceprint, retinal scans or other biological identifiers, along with some form of implant technology might serve to act as a mostly foolproof identification system. The system will have to be universal, however, or people may have identities on various networks—although this might not be a bad thing, depending on your view—and universal genetic or biological registration raises many privacy issues. It becomes more complicated if you want to identify an artificial intelligence, however, especially if that AI is not unique. It is possible that the network could take a “snapshot” of the AI’s code at some point in time and use that in place of a genetic database entry, or maybe use some variation of a public/private key system based upon the time of the AI’s registration with the network.
Either of these identification systems might be considered invasive to one’s privacy. Another method might involve registering the inhabitant’s DNA along with some anonymous identifier upon initial entry into the virtual world and only allowing a network login to that identity from the proper real person. This will only work properly if there is only one virtual world—once again a single person could have an identity in any world that is unconnected to the others where identities have been established.
Other science fiction-type identifications might be possible, such as brainwave patterns or non-transferable implants for real humans, or hardware keys for AIs (kept in a central repository.)
Once issues of identity are resolved, there needs to be a legal and/or physical way to determine the status of the spouse—is he alive or dead? Maybe the operative words here will have to be “active” or “inactive” rather than living or dead, since it may become really complicated to define life when AIs enter the picture.
In order to determine death, there must be some monitoring or registration system in place. One way to keep your real identity secret in the real world while maintaining a virtual self would be to leave instructions in your real world will specifying that, upon your death, a biological sample (or your network implant) must be sent to a repository along with medical proof of your death. The depository doesn’t know who you are, but has a biological sample matching your virtual identity. The doctor knows who you are, but does not include your name with the sample (or some intermediary strips the name from the sample as a matter of course.) This way, no one has all the information needed to tie your real self to your virtual self.
It may be that you have no need to keep your real and virtual selves separate, so in that case, the technician or doctor who removes your body from your VR interface equipment can report your death to the network when you die. If you die while outside of VR, then your will can specify, as part of finalizing your affairs, that the network needs to be notified of your death.
If you are in VR because you are alone in deep space somewhere, or if it is somehow impossible to determine or reach your physical location, the network must have some way of monitoring the condition of your ship or your equipment, and might report you dead upon the destruction of either.
If you are an AI, it is likely that you have an actual physical location (a server, ford instance). Your death could occur if your server goes offline and you do not have arrangements to reincorporate elsewhere (say, a backup server.) This will not work if AIs can move from server to server. In that case, the network will need to have some means of monitoring your “process”. If your process is killed—then you are dead—at least until you are restored from a backup. There will have to be some kind of reasonable time period for you to restore yourself or some means of verifying permanent death—but unless there is some physical piece of hardware that defines you as an AI, then there will always be some doubt as to your “death.”
Other ways of determining death might be to monitor network activity. The network might decide you are dead if you have no activity on the network for some period of time (akin to missing person’s law, where you are presumed dead if no one has heard from you for 7 or more years, for example). There might be some protections built into the system which put some of your assets into trust for some longer period of time, just in case you appear after a long absence (maybe a universal trust fund for people who fall though the cracks in the protection scheme for some reason.)
In any case, any report of your death should be announced on the network and in the real world so that you and people who know you have the opportunity to dispute your death, if necessary.
Once issues of identity and life status can be worked out, it remains to work out the mechanics of marriage in a virtual world. Getting a marriage license will probably require a biological sample, network process identifier, public key, or some combination thereof. The union can then be registered on the network, much like a real world marriage license becomes a public document. Once registration is complete, then whatever rights and responsibilities that accompany the marriage commitment will vest.
The best legal reason to register a marriage is so that the partners can obtain some legal benefit, such as health insurance (to care for your server or your real body), life insurance, and wealth or property transfer at death. This raises the issue of virtual property rights, which are beyond the scope of this paper, but should be at least mentioned in passing. There are two types of property that you might want to pass at death: network equipment and servers, and virtual property, including money and virtual real property. Real property, like servers, will probably be passed in the real world, unless you have it hidden from real world ownership conventions (say, by anonymous ownership which has been registered in a virtual world.) If, like in Neal Stephenson’s Snow Crash for instance, there is such a thing as valuable and permanent real estate, then you will want to control where it goes after your death. Maybe you don’t have any relatives in VR, or maybe you don’t want them to know that you have a virtual life as well as a real life.
If, like most people in the real world, you don’t have a will, you may still want to keep your virtual and real assets apart, especially if you have an “anonymous” virtual persona. If you have only a real world spouse, then maybe you want everything to go to him, but what if you have a virtual spouse (or children) as well? Along with virtual marriage comes virtual families and, quite possibly, virtual probate. Other issues raised by virtual families include child custody after divorce, which probably will mirror real world law—except in non-traditional marriages, where new rules will have to be developed.
Virtual marriage is the best way to perfect these virtual property rights and to care for your virtual family after death.
Marriage options in the real world are limited as of today to two partner male/female unions. There is no reason why this has to be the case in virtual worlds. In fact, some virtual worlds may be created entirely to escape the social limitations of the real world. There is also no reason that a partners in a virtual marriage have to both have a virtual presence (maybe one person has no interest in living in a virtual world, but at least some of the other person’s assets exist there—the marriage could be registered in VR such that the virtual intestacy scheme will not trump the real spouse’s rights in the virtual assets.
Ideally a virtual world would allow for line marriages, group marriages, unions of AI, non-human, and human partners, and open same-sex unions. Deep VR could allow for seamless interactions among disparate entities, without the physical/social complications that might arise in the real world.
Line marriages would be particularly useful in virtual worlds where it is impossible to produce virtual children. It also works in worlds where you can have children, but don’t want to dilute the holdings of your marriage by continually splitting and resplitting the shares of the property. In a line marriage, there are husbands and wives of various levels of seniority (depending on the time since joining), and new, usually younger spouses are invited in periodically to keep the line going. All property belongs to the marriage (each has an undivided interest in the whole) and not to their children (in a real world line-marriage—in this case there are no children). As long as one spouse is still alive, the property does not pass on to the children.
Group marriages (polygamy/polyandry) allow for maximum pooling of resources but suffer from the problem of marriage property being divided whenever a spouse dies.
A new twist on traditional and nontraditional marriages will occur whenever one or more of the partners in the marriage is not human, artificial, or formerly human. VR allows this, obviously, since the virtual projection of self does not have to reflect any actual physical characteristics of the original person, including life status (assuming there was an original person.) If an AI is sufficiently attractive emotionally to another being, whether human or AI, then physical appearance can be altered to suit the relationship.
It is likely that other types of marriage will be invented in new worlds, including maybe some forms of hive-mind, or “telepathic” gestalts.
The most interesting issue raised by virtual marriages is that of virtual children. The interesting questions arise, not when considering real children who can exist outside of VR, but children who only exist in VR.
Real children who cannot function outside of VR are only moderately interesting, since they have real bodies and will have genetic ties to real people. It is conceivable that a child could spend his entire life in VR, but laws analogous to the marriage laws discussed above would serve to protect them much like in the real world.
Of greater interest are “children” who are created in VR. A virtual child could be an artificial intelligence made from scratch, a digital copy of a real child, a digital copy of an adult modified somehow to be childlike, a fusion of one or more of the above, or an AI specifically designed to be childlike. In any of these cases, the virtual child has no existence outside of VR and is thus fully dependent on its parents, whether real or artificial.
Whether a child is an “off the shelf” model, or custom made using elements taken from real or artificial humans, there will probably need to be some rules on how to handle them, how to care for them. However, this discussion is only interesting if these children are considered alive and worthy of the protection of the law. If a virtual child is just a complicated simulation—basically a glorified version of a 1990’s era Tamagotchi (a virtual pet you kept on a key chain), then it seems pretty clear that there is no reason to worry about wealth transfer, slave labor or the right of the program to exist. If the program has legal rights as a living entity, then rules will need to be created in order to protect it.
The starting point for virtual children’s rights is some kind of registration system. When someone wants to create a child, he will have to petition to bring the child into existence and then register its process, hardware key, etc. (see discussion of life status above.) The virtual world administration might require that the prospective parent have a license to produce the child in the first place. A license might be based on merit, so “incompetent” parents will be filtered out at the outset. The parent will then have to assume parental responsibilities in order to gain parental rights. The network might have a mechanism for monitoring virtual children until (or if) they reach adulthood status. Alternately, each child might be required to have, as part of its interface with the network, a mechanism for reporting abuse. Like in the real world, abuse of a virtual child might be remedied by removing the child from its parent. Caring for abused virtual children might only consist of altering a portion of its code so that it becomes an adult and can take care of itself (commercial models will probably have this built-in), or allowing people who can’t afford to create virtual children to adopt. An interesting complication arises if the server that houses the child belongs to the person who originally created the child—making it necessary to have a mechanism to either confiscate the hardware or somehow move the child to another server. It might be that, as a prerequisite for creating the child in the first place, the hardware upon which the program runs must be physically located in a place where the parent has no physical control—that way legal authority will be able to transfer the child to another server without having to confiscate any physical property (other than a hardware key, if one exists) and will not run afoul of constitutional issues involving an unlawful taking. This option looks to be expensive for the governing authority, raising issues of how to fund the activity of government sponsored child protection. Note that this problem is minimized if an AI can be transferred between servers without any expenses beyond bandwidth, storage space, and processing power. Someone will still have to pay for it, but that might be taken care of by VR usage fees. Another solution might be to put the virtual child into storage—turning him off until someone decides that they want to create a child and is willing to pay the fees required to reactivate him. As an incentive, maybe the administrators of the virtual world will offer a substantial discount on reactivating a virtual child that has been placed in storage.
The whole idea of putting a child into storage raises another interesting issue—if you get tired of your kid, can you just put him into storage until you are ready to deal with him again? It depends. If the child has social interactions with other AIs, then prolonged deactivation might be cruel—the child might miss important events in the lives of others or return after a prolonged absence to find his childhood friends have outgrown him or his grandparents have died. At the other extreme, if the child knows no one outside of a single parent, then it is arguable that there is no harm in deactivation (beyond the child’s knowledge that he is not valuable enough to be kept running at all times.)
It might also be interesting to consider legal consequences of “killing” an AI child, but that issue is beyond the scope of this paper.
The next interesting issue regarding virtual children is numbers—should there be a limit? If there is a limit, what should it be? Say a person can afford to have ten thousand children and the network can support them. Say further that the parent wants all the children to be identical and uniformly bizarre looking and annoying, or worse, sporting Coca Cola emblems where their heads should be—leading to a proliferation of children that only a parent could love. Should the virtual world administration prevent this from happening? The problem might be even worse if no hardware is required to house the child (i.e. the child can exist on the network without its own server.) Without a registration system and parental responsibilities, this action might lead to “spamming” a virtual world with children. A solution might be to allow only one copy of a particular AI on the network at any one time. Further, we might require a parent to be an individual, rather than a corporation. A more practical solution might be to require a parent to be responsible for the actions of his progeny, thus providing a big disincentive to producing more children than you are capable of supervising.
Parental responsibilities might include tort responsibility, educational requirements, or bandwidth charges (programs require resources and resources cost money.) An alternate, but draconian, solution might be allowing the network police to deactivate any child left unsupervised for long periods at will. Popular opinion might be enough to keep parents in line—after all, who wants to be known as the guy who fathered a thousand annoying children?
One final issue regarding virtual children is growing up. If an AI is designed to be a child, it may never grow up—but what happens when the child’s parent dies or decides that he doesn’t want to care for a child anymore? As discussed above, it may be that a commercial child AI will have a software switch that causes it to become an adult whenever the parent wants it to do so. If a child is required to be licensed at birth, then it is possible that a requirement for licensing is a similar software switch. It is more interesting if both commercial and non commercial AIs simply grow and develop into adults—it is likely that most people who create AIs in the first place will want to experience the growth of a child AI into an adult. The problem here is that it is likely that child AIs will develop at different and highly unpredictable rates. There will have to be some kind of mechanism for determining at what point a child AI may become an adult. An interesting side issue is whether it might be legal to force an AI to remain a child against its will (or program it so it has no desire to grow up.)
One way of determining adulthood in AIs might be through some form of educational system—when the AI completes school it becomes an adult. Other ways might include adulthood exams or possibly an evaluation whenever the AI decides it is time to end its childhood.
Virtual marriage might be a useful institution in virtual worlds, but it probably will not replace real world marriage except in fairly extreme situations like in a world of AIs and human-to-software conversions or in the case of disasters that make actual reality unlivable. Of course, this assumption is made from the point of view of someone who has never experienced deep VR—maybe life in VR will be some much better than real life that it would be stupid to live anywhere else. If living in VR extends life or allows experiences that are so wonderful that actual reality pales in comparison, maybe it will be considered normal rather than extreme.
 See the case of a 600 lb man having heard problems and the trials of the crew trying to save him. “The Bravest” home page, available at http://www.hearstent.com/html/Bravest/bravest_Episodic1.html#600LB
 For a science fiction account of such a scenario, see Tad Williams’ Otherland series, where a huge computer network is built to house the personalities of the world’s richest people. See www.tadwilliams.com.
 For example, the character of Orlando, in the first book of the Otherland series, was born with a terminal disease and has never been able to leave his bed. Tad Williams, City of Golden Shadow, New American Library, 1998, ISBN# 0886777631.
 But note that the concept of ugliness may change dramatically once people become used to having perfect bodies in VR and begin to forget that real bodies are not perfect.
 See, for instance, the Joe Pantoliano’s character, Cypher, in The Matrix, Warner Brothers, 1999, sells out his friends for the chance to reenter a virtual world that seems, to him, far better than the real world. In one scene, he relishes an opportunity to eat a steak, something that he simply could not get in the real world for any price.
 See, for instance, Abre los Ojos (Open Your Eyes), Artisan Entertainment, 2001. See also, Vanilla Sky, Paramount Home Video, 2003, which is the American remake.
 See Robert E. Heinlein’s The Moon is a Harsh Mistress for the genesis of this idea. Berkeley Books, 1968.
 “Reports of my death have been greatly exaggerated” – Mark Twain.
 Neal Stephenson, Snow Crash, Roc Publishing, 1993.
 See endnote 7.
 Like the robot child in Steven Spielberg’s AI: Artificial Intelligence.
 Wait—isn’t this like the real world?