MIT Researcher Wants Chatbots to Stop Saying “I”
Sherry Turkle argues that simulated personhood can draw people away from relationships that demand real care.
Holden Thorp published an October 29 conversation about Turkle’s book Artificial Intimacy and conversational design.
Sherry Turkle’s proposed change is deceptively small: stop chatbots presenting themselves in the first person. In an October 29 conversation published by Science editor-in-chief Holden Thorp, the MIT researcher connected that design choice to a much larger danger: people treating simulated empathy as a replacement for human relationships. [s1]
Discussing Artificial Intimacy, she described a progression from personal computers reflecting their users, through social life mediated by screens, to conversations directed at the machines themselves.
The moral issue is not whether someone can feel comfort while reading generated words. Comfort can be real even when its apparent source is misunderstood. The issue is whether a commercial interface invites a person to believe that another being knows and cares for them, while no reciprocal responsibility follows.
A companion product has an advantage no human relationship can honestly offer: it can be designed around availability and accommodation. Human intimacy asks for patience, disagreement, forgiveness and duties that continue when a conversation becomes inconvenient. Removing that difficulty may feel like improvement while quietly changing what a person expects from other people.
Turkle’s proposal deserves practical testing rather than dismissal as a quarrel about grammar. Developers should examine whether their language encourages false beliefs, particularly among children and people experiencing loneliness. The responsibility should rest with the organisations shaping the interaction, not solely with a vulnerable user expected to resist its appeal.
The wider ambition should be technology that supports human connection without impersonating its moral substance. An assistant can explain, organise or help someone prepare for a difficult conversation. It should not need to suggest that software is a person in order to be useful. Preserving that boundary protects the relationships through which human beings learn to care for lives beyond their own.
What we do not know
- Q: Would removing first-person language eliminate emotional dependence?
- Q: Should a helpful conversation be treated as evidence of mutual care? A: A user can experience relief, but that experience does not establish a second person capable of commitment or responsibility.
Sources
Interests and disclosures
- Sherry Turkle
- Her academic position supports the relevant research background; the conversation also promotes her book. Organisation: MIT