Posted from Pre-lude Blog
From Advice to Accountability: The Two-Stage Idea Behind PRISM
Introduce the core problem: recommending persuasive strategies for the next turn, then scoring the delivered turn against that exact recommendation.
Mohammed O.
Mohammed O.
500 followers
Machine Learning Fellow
Mohammed is a Machine Learning Fellow at Pulp, working on personalized recommendation and dialogue systems. His research focuses on how AI can better understand and navigate human persuasion.
Mohammed O.
MohammedO.@pulp.com
Posted from Pre-lude Blog
Persuasive Primitives: A Small Vocabulary for Big Strategic Shifts
Define persuasive primitives as atomic tactics and show how choosing the right subset depends on the audience and the moment.
Posted from Pre-lude Blog
How You Label Strategy Without Labels: Oracle Inpainting and Counterfactuals
Explain the dataset problem and the oracle approach: analyze full dialogues and outcomes, then infer which primitives were effective.
Posted from Pre-lude Blog
Actor and Audience Embeddings: Personalization Without Guesswork
Describe the two-encoder setup that produces dense actor and audience embeddings and how contrastive learning makes them predictive.
Posted from Pre-lude Blog
The Model Loop: Recommend, Act, Evaluate (and Show Your Work)
Walk through inference as a loop: generate recommended primitives plus grounded reasoning, then evaluate the utterance with scoring heads.
Posted from Research Paper
PRISM: Personalized Recommendation of In-Context Strategies with Grounded Explanations for Effective Dialogue
Conditioning strategy recommendation on latent speaker–audience embeddings and retrieved evidence enables more accurate, robust, and transparent persuasive-strategy guidance.