Each upcoming event is scored by gpt-5.2 on five universal
dimensions (1–10) plus value for price. Universal means the scores
describe the event itself — the same for every viewer. The Score column
is the weighted composite (weights below).
Dimensions
- Social Quality
- Will attendees interact, or just sit in an audience? Bring-your-friends
events (Parks on Tap, concerts) stay moderate even if lively.
- Serendipity
- Crowd diversity + chance of an unexpected encounter. Outdoor public events
and walk-up third spaces score higher; private ticketed seating lower.
- Uniqueness
- Rare one-time event vs. something that happens regularly?
- Community Building
- Does attending build something durable? Worship/intra-congregation events
cap at 5; secular societies, clubs, and cultural events at religious venues
are not capped.
- Solo-friendliness
- Can a stranger show up alone and meet people? 1–3 bring friends,
4–6 neutral, 7–10 built for strangers to meet (clubs, civic meetings,
volunteer days, classes with regulars).
- Value for Price
- Is it priced fairly for what you get? Now a small part of the composite.
Composite weights
Description quality flag
The D column is the LLM's read on source text:
● rich,
● thin,
● missing (title only — take scores
with a grain of salt).