Milton data projectThe AI said Milton had no active playgrounds. The counts said otherwise
The Town of Milton publishes a Park Amenities layer as open data: one row per sports field, court, playground or splash pad, with a STATUS column. We asked a simple question of it, wrote the kind of filter an assistant produces first, and then checked it. Script and figures are our own work, from a copy of the file taken on 23 October 2026.
The question a learner types: “Write Python that lists the active playgrounds and splash pads in Milton.” A typical first draft filters on STATUS == "ACTIVE". It runs without an error, and for playgrounds it returns nothing.
Our count of the Town of Milton Park Amenities file (166 rows, downloaded 23 October 2026): the draft filter against the checked one| Category | Rows in the file | Draft filter finds | Checked filter finds |
|---|
| PLAYGROUND | 13 | 0 | 13 |
| SPLASH PAD | 12 | 11 | 12 |
| BASKETBALL | 22 | 14 | 20 |
| TENNIS | 15 | 15 | 15 |
| PICKLEBALL | 3 | 0 | 3 |
Why? The STATUS column holds the same word in two spellings. Our count: 120 rows say ACTIVE, 44 say Active, and 2 say Historic. Every playground in the file is spelled Active, so an exact match on the upper-case word skips all 13. The draft is not broken in a way Python can see. It is wrong in a way only a person who looked at the data can see.
The whole exercise in one standard-library Python file, run next to the downloaded CSV:# Milton park filter, draft versus checked: our teaching example, not Town code
import csv
rows = list(csv.DictReader(open("park_amenities_hub.csv", encoding="utf-8-sig")))
def draft_active(category):
"""The first draft, the kind an assistant writes: exact match on STATUS."""
return [r for r in rows if r["CATEGORY"] == category and r["STATUS"] == "ACTIVE"]
def checked_active(category):
"""After the human check: compare STATUS without caring about case or spaces."""
return [r for r in rows if r["CATEGORY"] == category and r["STATUS"].strip().upper() == "ACTIVE"]
# Human check 1: every row must land in exactly one STATUS bucket.
buckets = {}
for r in rows:
buckets[r["STATUS"]] = buckets.get(r["STATUS"], 0) + 1
print("rows", len(rows), "status buckets", buckets, "sum", sum(buckets.values()))
# Human check 2: the draft and the checked filter must agree, category by category.
for cat in ["PLAYGROUND", "SPLASH PAD", "BASKETBALL", "TENNIS", "PICKLEBALL"]:
total = sum(1 for r in rows if r["CATEGORY"] == cat)
print(cat, "in file", total, "draft", len(draft_active(cat)), "checked", len(checked_active(cat)))
Check 1: buckets that add up
Count every distinct STATUS value and add the counts. They must come to 166, the number of rows. Seeing ACTIVE and Active listed as two buckets is the moment the bug becomes visible, before any filter is written.
Check 2: draft against fix
Run the hasty filter and the careful one side by side, category by category. Where they agree, as for tennis, nothing is proven; where they differ, as for playgrounds, a decision is needed about what the data means.
Check 3: read the two odd rows
The two Historic rows are OBJECTID 162 and 163, both a half basketball court at Community Park. The Town's description says the layer holds active amenities, so these two are worth a question to the publisher rather than a silent guess.
One more variant the counts reveal
Grouping by address splits one site in two: our count finds 805 SANTA MARIA BLVD on 21 rows and 805 SANTA MARIA BOULEVARD on 8. An agent asked how many amenities sit at that address would answer 21 or 8, both wrong, unless someone normalises the street type first. The file has no date column, so it cannot tell you when any amenity opened or closed.
The layer is the Town of Milton's own, described by the Town as updated on a daily schedule, so counts on a later download may differ from ours. Used under the Open Government Licence - Milton, which permits use including for commercial purposes (terms of use). Contains information licensed under the Open Government Licence - Milton. The Town does not endorse this page, and we use no Town crest or logo.
