genus perspective
Where does an AI conversation lead next?
An AI conversation may help someone explore a next step, but time spent chatting is not the whole story. An early study prompts careful questions about personal purpose, human connection, and useful choices beyond the screen.
What we found in the news
A March 21, 2025 preprint from MIT Media Lab and OpenAI (opens in a new tab) examined chatbot use and psychosocial experiences over four weeks. The authors called for more research on supporting people:
“without fostering dependence or replacing human relationships”
— Cathy Mengying Fang and colleagues
The study compared chatbot conditions, without a group that used no chatbot. Associations involving people’s chosen usage time do not establish that heavier use caused the reported experiences. The original release invites careful questions, rather than a verdict that conversational AI is uniformly helpful or harmful.
Why this matters to us
A conversation with a tool may help someone find words, explore an interest, or organize a question. The next step in that person’s life matters too.
Imagine an adult using an optional assistant to think through joining a local walking group. They find a suitable public invitation, set the tablet down, and call a friend about going together.
That is a possible use, not an outcome we are claiming the research demonstrated. The person might instead decide the group is unsuitable, ask the organizer a question, or choose a different activity.
At genus, we are interested in technology that supports people’s own connections and purposes. Time spent in a conversation tells us little about whether those purposes are being served.
Ask what the person hoped to do
Begin with the person’s reason for using the tool. They may want practical information, a way to prepare for a conversation, or help exploring an unfamiliar option.
Avoid turning that purpose into a target someone else has chosen. Joining more groups is not necessarily the right goal for every person. Rest, a familiar routine, or a quieter form of contact may be what they want.
In our imagined walking example, the useful question is whether the information helped the person decide what to do. The assistant’s number of replies is secondary.
Give the person room to say that the tool did not help. They can know that an answer missed their concern even when it was friendly and well written.
Make the route beyond the screen practical
If a service points someone toward a group or resource, confirm the information people need to act: the organizer, the meeting place, access details, and how to ask a question.
A suggested activity that is unavailable or difficult to reach can leave the person with more work. Current local information and a real contact may matter more than another conversational turn.
Offer choices without implying an obligation to follow a suggestion. The person remains free to close the tool, speak with someone they trust, or decide later.
Be clear about what the assistant is and who provides the service around it. A warm conversational style should sit alongside an accurate explanation of the tool’s role and the available human contact.
Review usefulness without reducing a person to a usage count
When someone chooses to give feedback, ask about the task and their experience. Was the information understandable? Did they find a useful next step? Was anything confusing or difficult to stop?
Handle that feedback with appropriate privacy and a clear purpose. A service does not need unrestricted access to someone’s personal conversations to ask whether its resource guidance was helpful.
Look at patterns with care. Frequent use may have several explanations, and a brief visit can still accomplish the person’s purpose. Avoid making assumptions about someone’s emotional state from activity data alone.
Use what people tell the team to improve the design: the available choices, the clarity of the role, or the route to people and activities they value.
What this means for the people you serve
People can decide what kind of support they want from a tool and whether to use it at all. Their relationships, preferences, and offline life remain the context. A useful experience helps them pursue something that matters to them, with room to stop or choose another route.
The genus connection
Today, genus care-app and care-wall experiences can connect configured resources, activities, and communication with supported guidance. The organization defines the available content, access, and role of AI in that setup.
We would begin with the person’s intended next step and keep human connection in the service design. Clear local information, suitable invitations, and an understandable way to reach someone can make the technology part of a wider support experience.
One question to take away
After this conversation with the tool, is the person better able to pursue the connection or activity they chose?
Supporting references
- How AI and Human Behaviors Shape Psychosocial Effects of Chatbot Use: A Longitudinal Randomized Controlled Study (opens reference preview; direct link opens in a new tab)
MIT Media Lab and OpenAI · 2025-03-21
The original March 2025 preprint examines chatbot conditions and reported psychosocial experiences. Its design and observational usage findings do not establish universal causal effects.
