WEBVTT

00:00:00.700 --> 00:00:04.810
Who studies alternate wetting and drying
in rice, where, and with whom?

00:00:05.462 --> 00:00:09.250
One request, and an agent maps
the whole field from OpenAlex.

00:00:10.682 --> 00:00:15.168
Then a second agent checks every
number, before anyone sees the page.

00:00:17.200 --> 00:00:20.447
AI for Research Efficiency.

00:00:20.497 --> 00:00:23.662
Episode 11:
map a research field with OpenAlex.

00:00:26.700 --> 00:00:30.293
Last time, an agent fetched
research data through APIs,

00:00:30.384 --> 00:00:33.052
starting with OpenAlex,
which needs no key.

00:00:33.809 --> 00:00:38.461
This time, one real run, from a
question to a checked map and a page:

00:00:38.893 --> 00:00:43.283
write the rules, let it estimate and
fetch, and check before you share.

00:00:46.100 --> 00:00:48.167
The rules go in an instruction file,

00:00:48.217 --> 00:00:52.209
as in episode 3: the question,
the source, and a budget.

00:00:53.049 --> 00:00:56.293
Without a key, OpenAlex gives
you 1,000 credits a day.

00:00:56.764 --> 00:01:00.419
A page of results costs 1;
a search costs 10.

00:01:00.870 --> 00:01:05.976
So the file says: stay under 300, and
estimate before paging through anything.

00:01:07.076 --> 00:01:09.059
It asks for scripts that re-run,

00:01:09.169 --> 00:01:12.353
and a log of every address it calls,
with what it cost.

00:01:13.321 --> 00:01:16.168
And it lists what to make:
works per year,

00:01:16.359 --> 00:01:20.730
the leading institutions and journals,
and who writes with whom in Arkansas.

00:01:24.500 --> 00:01:26.745
Then the request, and the run begins.

00:01:27.436 --> 00:01:31.725
The clock is the run's own:
7½ minutes, shown faster.

00:01:32.936 --> 00:01:35.522
First it reads OpenAlex's
guide for agents.

00:01:35.913 --> 00:01:39.230
Then it writes one small module
that every script shares:

00:01:39.561 --> 00:01:42.266
it logs each call,
and keeps each answer.

00:01:46.100 --> 00:01:50.208
Then one search defines the field:
alternate wetting and drying,

00:01:50.339 --> 00:01:55.588
or intermittent irrigation, or AWD,
together with rice or paddy.

00:01:56.746 --> 00:01:59.796
It finds 2,945 works.

00:02:00.268 --> 00:02:03.237
Before going on,
it reads 20 of them at random.

00:02:04.770 --> 00:02:07.053
All 20 are about rice and its water.

00:02:07.384 --> 00:02:11.639
About 13 are squarely on the practice,
and about 7 sit at the edge.

00:02:12.110 --> 00:02:15.114
So the search finds the field,
with a margin around it.

00:02:17.200 --> 00:02:20.364
Every answer states its cost,
so it counts as it goes.

00:02:20.834 --> 00:02:21.925
The counts by year,

00:02:22.036 --> 00:02:27.061
by institution and by journal are one
grouped search each: 10 credits apiece.

00:02:28.212 --> 00:02:30.558
A wrong field name
was refused for free.

00:02:30.969 --> 00:02:32.684
And to confirm each institution,

00:02:32.794 --> 00:02:35.480
it fetched single records,
which cost nothing.

00:02:38.900 --> 00:02:41.765
Works per year: 16 in 2000,

00:02:41.996 --> 00:02:46.283
65 in 2010, and 325 last year.

00:02:47.843 --> 00:02:49.765
The institutions that publish most:

00:02:50.156 --> 00:02:54.272
the International Rice Research Institute
leads, with 138.

00:02:54.723 --> 00:02:58.147
UC Davis is the only one in
the United States on the list.

00:02:59.447 --> 00:03:00.269
And the journals:

00:03:00.721 --> 00:03:04.450
Agricultural Water Management,
then Field Crops Research.

00:03:07.700 --> 00:03:10.306
For Arkansas,
it never filters by a name.

00:03:10.597 --> 00:03:13.654
It looks up each institution's
OpenAlex number first:

00:03:14.005 --> 00:03:17.673
11 records, from Fayetteville
to Stuttgart and Jonesboro.

00:03:18.993 --> 00:03:23.824
44 works have an author there:
1.5% of the field. 126 authors,

00:03:23.874 --> 00:03:27.121
and 501 pairs.

00:03:29.122 --> 00:03:32.784
Each dot is an author,
each line a paper written together.

00:03:33.174 --> 00:03:38.046
The agent sees two groups: the university
with the USDA unit in Jonesboro,

00:03:38.216 --> 00:03:41.937
on methane and water, and the
rice research center in Stuttgart.

00:03:46.000 --> 00:03:49.079
Before any check,
it writes down what looked wrong.

00:03:49.960 --> 00:03:53.184
Some researchers have two
or three OpenAlex profiles,

00:03:53.274 --> 00:03:54.726
so they appear more than once.

00:03:55.157 --> 00:03:57.490
One paper lists its
four authors twice,

00:03:57.720 --> 00:04:01.104
and the author with the Arkansas
address has no profile at all.

00:04:02.504 --> 00:04:04.186
And the last few years jump.

00:04:04.557 --> 00:04:08.371
The agent notes that more datasets
and repository records may be part of it,

00:04:08.542 --> 00:04:11.405
so recent years aren't comparable
with earlier ones.

00:04:14.900 --> 00:04:19.309
Then the checker: a second agent,
with its own script and its own calls,

00:04:19.360 --> 00:04:21.583
that doesn't reuse
the first one's code.

00:04:22.724 --> 00:04:28.007
Every count matched: 26 years,
16 institutions, 11 journals,

00:04:28.117 --> 00:04:33.019
the 44 works, and all 501 pairs,
rebuilt from scratch.

00:04:34.615 --> 00:04:39.424
It found one thing: the yearly counts
add up to 25 fewer than the total.

00:04:39.755 --> 00:04:44.182
Most likely, those works have no year.
The agent added that to its notes.

00:04:46.100 --> 00:04:49.793
Only then the page: four figures,
each with its query,

00:04:49.985 --> 00:04:52.673
the date of the data,
and the addresses to check it.

00:04:54.158 --> 00:04:55.499
It's on your computer for now.

00:04:55.750 --> 00:04:56.361
To share it,

00:04:56.411 --> 00:05:01.055
put the folder on GitHub and publish it
with Vercel, as in episodes 5 and 6.

00:05:05.200 --> 00:05:08.863
Read the map for what it is.
One search defines the field.

00:05:09.113 --> 00:05:12.746
The counts are OpenAlex's matches,
not a ranking of quality.

00:05:13.056 --> 00:05:14.957
And check any name you quote.

00:05:16.312 --> 00:05:21.319
The whole run: 7½ minutes, 47 calls,
and 112 credits of the day's 1,000.

00:05:21.369 --> 00:05:25.014
A free key gives 10 times as many.

00:05:26.800 --> 00:05:30.807
Pause here and try it: copy the
instruction file from the companion page,

00:05:30.938 --> 00:05:32.711
put in your own question and field,

00:05:32.821 --> 00:05:35.706
and read the checker's report
before you make the page.

00:05:38.800 --> 00:05:43.127
So: define the field with one search,
estimate before you page,

00:05:43.318 --> 00:05:46.663
keep the scripts, and check
every number before you share.

00:05:48.400 --> 00:05:51.602
Next: make a narrated
explainer of your paper.