People replicas

Promises

NLP systems allow to train personalised chatbots that capture personality traits of a deceased or living person.
NLP can be used to create digital replicas that imitate the speech and language patterns of deceased individuals. A chatbot is able to converse by imitating a deceased individual via a learning process based on conversational data collected from this person. Typically, such chatbots do not repeat the training data but generate new phrases that the imitated person has never uttered, but that are very similar in style.

A conversational agent that imitates a deceased individual would most often be used by someone who knew the person while they were alive. The user can enact their wish to remember the deceased by embracing the illusion of their presence.
There are already ongoing projects replicating famous philosophers or authors and generate “new” texts which they would write if they were still alive. Also, with a chatbot called “deadbot” or “griefbot”, people can “talk” with a deceased family or friend.
Additionally, NLP can be used to have a virtual friend. Without physical interactions and social obligations, people can enjoy conversations with a chatbot as if they would feel like talking with a real human. The virtual friend is a personalised system so that it can intercept the people’s feelings from the responses and answer what the people need.

The most recent chatbots raise ethical questions that relate to the use of affective computing – a range of techniques that help perceive and simulate emotions that influence user behaviour.
In the long term, the effects of chatbots, including “deadbots” or “griefbots”, may produce a significant change in the human condition through the co-adaptation of language between human users and conversational agents.

Respect for the memory and dignity of the dead is a widely shared principle and it is questionable whether the development of “deadbots” should be forbidden or regulated by legal measures. In the latter case, a specific legal framework along with technical constraints limiting the side effects on the natural mourning process must be devised.
Is there a consideration for posthumous personal data treatment, including images of individuals?
A “deadbot” is able to generate original outputs that the person they are imitating never uttered during their lifetime. Such applications may do reputational damage or otherwise infringe on the person’s dignity after their death. How can we ensure that this does not happen?
What options are presented to the data subjects to have control over what happens with their data posthumously?