An experiment platform designed for reproducibility.

For behavioral economists, social scientists, and academic research groups. Design treatments, fund them on-chain, and export the full run as a verifiable record.

Sign up as Experimenter

experiment.yaml draft

name: replication_v2
template: trust_v1
participants: 60
variations:
  - { E1: 10, m: 3 }
  - { E1: 20, m: 2 }
  - { E1: 10, m: 5 }
  - { E1: 20, m: 5 }
funding: <escrow>
export: csv | json

i. Design

Designing your experiment.

Pick a template, fix base parameters, declare which parameters to vary. The platform generates factorial combinations from your varied values and deploys each variation as its own contract. Everything else stays at base.

design summary

Base parameters: set
Variation matrix: 4 (2 × 2)
Max matches / variation: 15
Assignment: round-robin

ii. Deploy

Running it.

Open registration when you're ready. Participants self-enroll and are assigned to variations automatically. Monitor enrollment and match progress in real time as decisions land on the chain.

match status live

v1	●	14 running	12 settled
v2	●	11 running	13 settled
v3	◐	6 running	8 settled
v4	○	0 running	0 settled

iii. Fund

Funding, escrow-first.

Each variation is funded up front by the experimenter through escrow held in the smart contract. The escrow covers the worst-case payout for every pair; anything that isn't distributed is refunded when the experiment ends. The experimenter never holds participant funds.

escrow model

escrow = worst-case payout × number of pairs

payouts = what actually distributes

refund = escrow − payouts

iv. Export

The data you get back.

Every enrollment, decision, and payout is recorded on the blockchain. Pull a run as CSV for analysis or as signed JSON for audit and replication. The manifest hash is part of the record, so anyone can verify the experiment's design from the data alone.

run.csv for analysis

match_id, variation, i_sent, t_returned
001, v1,  6,  12
002, v1, 10,   0
003, v2,  4,  10
004, v2,  0,   0

run.json for audit

{
  "manifest": { ... },
  "variations": [ ... ],
  "matches":    [ ... ],
  "signature":  "0x..."
}

Run the
experiment.