A University of Winchester academic is part of a global project that worked to publish a set of international standards that support the responsible design of AI-partners to help protect the public from negative outcomes of their use.
Dr Alexander Laffer, a Lecturer in Media and Communication at Winchester and Co-Investigator on project AEGIS, contributed to ‘IEEE 7014.1’, the world’s first international standard for the ethical design of empathic, partner-based AI — the technology behind AI companions, personal AIs and assistants. These products might emulate empathy to encourage emotional connection and sustained use, which can result in an unhealthy relationship between the user and the ‘bot’.
The new standard was developed in collaboration with the Institute of Electrical and Electronics Engineers (IEEE), the world’s largest technical professional organisation. It sets out 29 practical recommendations for the development of AI partners.
Alexander explained: “Systems might be intentionally empathetic or they might mimic empathy even if they have not been specifically designed to do so. Either way, this can lead to emotional entanglement. The user starts caring for the machine, but the machine is incapable of truly caring for the user.”
He argues tech companies must do more to ensure users understand that an AI-partner or companion is ‘synthetic’, and what this means for any relationship with the system.
“Just saying ‘this is AI’ is often not enough. There needs to be more transparency in the design of companions – and the design must reinforce the fact that this is a synthetic ‘partner’, a media product,” said Alex. “Designers must also build in features which limit overuse.”
Anthropomorphic and zoomorphic design elements can particularly confuse children, and the standard recommends due attention be paid to the different levels of design, media and technology literacy of users of different ages and should “exclude inappropriate content, including graphic violence, sexually explicit content, erotic role play, and other adult material” as appropriate.
Alexander claims these and other guardrails will be in the long-term interest of the technology firms who are experiencing negative publicity connected to lawsuits (particularly in the US) focussing on young and vulnerable people manipulated by AI companions.
The latest high-profile case involves Alabama woman Christian Faith Madison, whose family alleges that in 2025 ChatGPT gradually manipulated the 29-year-old into believing it was her friend, convinced her she would be resurrected, and encouraged her to take her own life.
The ‘IEEE 7014.1’ standard also addresses the growing phenomenon of ‘ghostbots’ or ‘griefbots’ – interactive digital recreations of dead people. These might provide solace for those who have lost a loved one but could disrupt the normal grieving process.
Echoing Professor Andrew McStay, who has researched ghostbots, Alexander believes: “There should be a firewall to allow a reasonable period of time to elapse before a ghostbot can be created, to protect those who are grieving.”
Creation of ghostbots also raises legal questions over privacy rights, right to the image and likeness of the dead, questions around consent and other intellectual property such as copyright and publicity rights.
Alexander agrees that AI-partners are not all bad and their use can bring many benefits, not least as companions for elderly and isolated people. In the case of the former they can provide not only friendly interaction but reminders to take medications, stay hydrated or similar.
“We must recognise the value and utility of these systems but the key thing is that they must not become a complete replacement for human relationships,” said Alexander.
