Robot nannies and applied ethics

In the journal "Science", 24 April 2009, pp. 463-464, two letters from computer specialists take different sides on the desirability of proposed "child-minding" robots. One letter states, "Intuitively, we... suspect that nanny-bots are not good for the psychological development of children left in their care, but until empirical research demonstrates this, we must suspend judgment; such research might, in fact, find no harm at all." The other letter rejoins, " I think that waiting for empirical research to demonstrate psychological harm to children is dangerous. Suspending judgment about possible harm when many empirical studies show the lasting effects of neglect is not a good option."

A critical point about this disagreement has to do with the fact that there will be no empirical research on "nanny-bots", unless some dramatic changes in research ethics take place. Research that has the potential for harming children is not approved, much less funded, unless its dangers are balanced by the likelihood of finding information that will prevent some existing cause of harm from functioning. "Primum non nocere"-- first, do no harm-- is the principal guideline for researchers who work with infants and children. Institutional Review Boards, groups that review and approve or disapprove research plans, are equally concerned with the possibility that experimental work might cause developmental problems.

Very well, though, aren't there different approaches to research? Nanny-bots are not against the law. Suppose some parents chose to use them? Could researchers not collect data about the bot-reared children and use that information to decide whether robots fostered good development? Yes, this could be done; but no, the results of this research would not give us satisfactory answers about the effect or robot-rearing. This type of quasi-experimental research design is plagued by confounded variables-- factors that occur together in such a way that we can't figure out which of them has which effect. For example, we might guess that the average parent would not choose to use a nanny-bot. What kind of parent might make this choice? We can only speculate, but such a choice might accompany an unusual level or type of education; family income; family realtionships outside the marriage; marriage itself; involvement with the community; professional demands on the parents; health status of parents; mental health and personality; characteristics of the child that make him or her difficult or easy to care for; and many other possibilities. If a bot-reared child develops worse than other children-- or better-- we have no way to know which factor or factors caused the difference.

Interestingly, the authors of these letters were concerned that robot-rearing might be analogized to neglect, and have effects similar to neglect, but presumably a nanny-bot would be built to respond quickly to infant cues. The big problem would seem to be that the nature of those cues changes rapidly in the course of early development, so the nanny-bot would need to be re-set to different types of cues as time passed. The authors also referred to "bonding" and "attachment" as problems resulting from the children's "anthropomorphic projection", but the meaning of this concern was not clear. The most likely developmental problem for the bot-reared would probably have to do with communication and language, both of which involve complex cognitive abilities on the part of adults partnering with communicating infants.

Goodman, K.W., & Einspruch, N.G. (2009, April 24). The way forward in the world of robotics. Science, 324, 463-464.

Sharkey, N. (2009, April 24). Response. Science, 324, 464.