What we offer
We provide services to campaigns and organizations whose goals are clearly aligned with public good. We do not develop or test messages that are false, misleading, or that incite exclusionary attitudes or violence. If you’re interested in working with us, reach out to team@rhetorical.org
Crowdvocate.org is our platform for crowdsourced messaging. Crowdvocate helps organizations with exploratory message development, by hosting tournaments where diverse users compete to write novel and high-impact messages.
To sponsor a tournament, a client defines the target audience and outcome (e.g. attitude or belief), and then our panel of hundreds of users attempt to design the most impactful message for moving the outcome. We efficiently test the messages using adaptive randomized experiments, award prizes to the winning users, and share all results with the sponsoring organization.
AI models
We develop computational models to predict the impact of a given message on public attitudes and beliefs, through meta-analysis, LLM-based simulation, and fine-tuning of domain-specific models from experimental results.
We offer access to our predictive models through an API, which can be used to receive fast, AI-based feedback on messaging. To request access, contact team@rhetorical.org
Consulting
We are academic researchers with extensive expertise in the design and analysis of randomized survey experiments to measure the effects of messaging on public attitudes and beliefs. We have particular expertise in bespoke and technical tools such as adaptive experiments comparing a large number of messages (100+), identification of heterogeneous effects for microtargeted messaging, measurement of effect persistence through multi-wave panel experiments, and experiments involving AI-generated or factorial treatments.
For less bespoke projects such as rapid message tests comparing a set of hand-designed messages or videos via randomized survey experiments, we suggest organizations work with Swayable, Blue Rose Research or GrowProgress.