for researchers

Screen studies for a review

Sort the clear studies in or out, and hand a person the ones that give too little to judge. decide with a band does both.

The survey with a result is true and the opinion essay is false. The bare title gives too little to judge. Inside the band 0.1:0.9, it is not sure. It prints null, and a person reads it.

question="Does this report measured results on remote work?"

cat <<'EOF' |
Survey of 900 remote workers: output rose 4%.
Opinion essay on open-plan offices.
Title only: Remote work and output.
EOF
thinkthen decide "$question" \
  --batch 1 \
  --lines \
  --threshold 0.1:0.9 |
jq .

Output

{
  "input": "Survey of 900 remote workers: output rose 4%.",
  "value": true
}
{
  "input": "Opinion essay on open-plan offices.",
  "value": false
}
{
  "input": "Title only: Remote work and output.",
  "value": null
}

exit 0

The functions it uses

decideAnswer one yes or no question about the evidence.

On GitHub: github.com/botassembly/thinkthen