The Festival List That Kept Forgetting Quiet Films
In the festival office, a programmer hit play and a sorter snapped thousands of film entries into a ranked list. A volunteer frowned. Last year a quiet small-town short stole the show, but this year those voices barely appeared. The team joked, “Hiring is like this list, just with people.”
They traced the path: who even hears the call for films, who gets filtered by quick checks, who reaches the jury, and what reviews say after the screenings. Then the loop: next year’s sorter gets tuned using last year’s crowd reactions. If the crowd missed someone, the tool learns to miss them again.
The coordinator started a messy list of tilts. Some were built into the place: prestige habits, friend networks, and who can afford to keep making work. Some were life: travel limits, caregiving gaps, avoiding unsafe spaces. Some were tool blind spots: ads shown unevenly, clunky portals, video tools that read some faces or accents better, and people trusting a score too much.
They realized they’d been asking different “fair” questions. Is the invite list balanced? Are the scores equally right? Are some films pushed so low they never get watched? Is the process respectful? A simple acceptance ratio could look fine while the top of the ranking stays narrow, year after year.
Fixes came in three flavors. Before sorting, they could change the call and examples so the sorter doesn’t learn old habits. While tuning, they could make it lean less on easy identity hints like names, faces, or accents. After sorting, they could reshuffle the list to meet a representation goal, but that can require knowing private traits at decision time.
Then a new worry: they had lots of neat records from early steps, but little they trusted about what happened later. Did invited filmmakers feel respected, earn anything, or keep working? Their info also leaned toward a few languages and places, and it often missed disability or forced gender into a small set of boxes. A sudden travel shock could flip everything.
On the final night, the coordinator stopped staring at the sorter like it was the whole story. The real job was watching the whole pipeline, who hears, who gets cut early, who shows up near the top, and what feedback trains next year. Tools that judge people from video and faces felt like the riskiest shortcut, so they chose extra restraint.