Interaction conditions were randomized, but voluntary usage duration was not. The study did not compare chatbot use with no chatbot use, and assigned conditions did not show significant overall effects on the principal outcomes.
What the study can and cannot tell us
A four-week study with 981 completing participants examined chatbot interaction modes and conversation types, alongside patterns of voluntary use. The paper reports associations between heavier use and worse outcomes on measures including emotional dependence and problematic use.
That does not establish that longer use caused those outcomes. Usage duration was not randomized, and there was no non-chatbot comparison group. The assigned experimental conditions also did not show significant overall effects on the main outcomes. The uncertainty belongs beside the finding.
Design the work around people and outcomes
For a business, an assistant should help people accomplish defined work and collaborate effectively, not become the only place they seek judgment or support. Clear tasks, human review, and real team relationships remain valuable.
- Set a useful outcome rather than equating more conversation with more value.
- Keep consequential decisions connected to accountable people.
- Avoid presenting a conversational product as a substitute for personal or professional support.
A PRACTICAL THADDEUS WORKFLOW
Give the assistant a task and the team a shared context.
Collab helps teammates work from common project material and clear ownership. Otto can assist with supported browser tasks under your direction. Position both around practical work and human relationships, rather than promises about emotional wellbeing.
Try this assignment
Prepare a concise project handoff for our team. Identify completed work, open questions, and the decisions our named reviewers need to make. Keep the focus on the assignment and do not invent agreement or personal commitments from teammates.
This is not a claim that Thaddeus prevents emotional dependence, improves mental health, or is a safer companion than the products studied. Otto and Thaddeus are not substitutes for human relationships or qualified support.
Sources & further reading
- How AI and Human Behaviors Shape Psychosocial Effects of Chatbot Use ↗
Research paper, arXiv 2503.17473. 2025 paper; reviewed October 2, 2026. - Readable mirror of the chatbot-use research paper ↗
Hugging Face Papers. Paper 2503.17473; reviewed October 2, 2026.
Educational information, not a security audit, professional advice, or guarantee. The assignment is an example to evaluate, not a completed demonstration. Historical incidents and product capabilities do not establish comparative safety. Confirm current policies and agreements for your intended workflow.