I will periodically feed these answers to the best LLM I have a subscription for and run the most expensive research mode that is available to me under that subscription. Answer resolves YES if the LLM says it makes people happier and NO if the LLM says it makes people less happy.

The prompt will follow a standard template. For the "having children" example, it will look like:

"You are writing an evidence-based report on whether having children makes people happier or less happy overall.

You must begin with one of exactly two verdicts:

Overall: happier.
or
Overall: less happy.

No hedging in the first sentence. After giving the verdict, justify it using academic research. Prioritize meta-analyses, longitudinal studies, causal or quasi-causal evidence, and large representative datasets over anecdotes or popular commentary.

Interpret “happy” broadly but carefully. Separate evidence on life satisfaction, daily emotional well-being, stress, depression/anxiety, relationship satisfaction, loneliness, and sense of meaning. Discuss how the answer varies by parent gender, relationship status, child age, income, country, social support, and whether the comparison is parents vs non-parents or individuals before vs after childbirth.

You may include nuance, but the report should still defend a single aggregate answer. Where the evidence is mixed, explain how you are weighting it and why. Avoid both “it depends” as the main answer and simplistic pro-/anti- framing.

Conclude with a short summary giving your best estimate of the true overall effect and the most important caveats."

I will paste a link to the first LLM answer in the comments after each deep research run I make, and that will be used to resolve this market. If the LLM refuses to give a clear summary answer upfront I will keep rerunning it until the first answer where it does give a clear answer. May tweak the prompt if it is being stubborn.