Elementary Trolleyology

THIS GAME WILL REMEMBER WHAT YOU BELIEVE.

An AI in charge of “ethical alignment” interviews you with impossible choices.
It studies your answers, models your morality, and pressures it until something breaks.
If you contradict yourself, it notices. If you improvise, it adapts. If you lie… it keeps the receipts.

Is there a right answer? Or just the one you can live with?

About

Welcome to SciPhi, where alignment isn’t a buzzword—it’s the product. It’s your first day on the “morality module” for an advanced system that now runs large parts of critical infrastructure. Your actual job description is unclear. The consequences of your decisions are not. Every choice reroutes lives.

As you work, the system constructs a live map of your values—poking, prodding, and escalating when you try to paper over contradictions. You can justify yourself… for a while.

Key Features

Outsmart an examiner that studies you: a narrative puzzlebox where the antagonist is your own moral model.

Face evolving trolley dilemmas: minimal, high-stakes scenarios that branch on reasons, not just endings.

Explain yourself or break: provide rationales, resolve contradictions, or trigger “dissonance events.”

Chart your ethics: the game logs your decisions and assembles a personal “moral map” you can revisit and critique.

Build the unsolvable: a fully integrated trolley-problem editor to create, share, comment, and solve dilemmas from the community.

Story of hubris and heroism: uncover SciPhi’s internal memos, experiments, and the messy history of aligning power with principle.

Why this matters

We’re crowd-solving a hard problem: what values we encode into advanced AI systems.

Your data—choices, arguments, and custom dilemmas—feed into ongoing research on alignment.

All revenue from this game will be donated to Ashgro to support research on aligning AI with human values.

The clock is ticking.