Cognir Ink.

Cognir's FEMME Simulator

The FEMME simulator is a heavily guard-railed prototype created for the founder of FEMME, who has regarded it positively and integrated it into their systems, which are, at the time of writing, under development. The collaboration with FEMME remains active and ongoing.

What is FEMME?
FEMME is a high school organization based in Ontario, Canada. The founder, Silin Cai, is a student at Abbey Park High School. Their mission is to create and improve situational awareness and develop safeguarding skills.

What it is: The FEMME Simulator is an educational training tool that helps women and girls practice recognizing and responding to uncomfortable or unwanted social interactions. It places the user inside realistic scenarios, such as being approached at a bus stop, followed in a parking garage, pressured at a party, or put in an uncomfortable workplace situation.

The simulator uses an AI character to roleplay the interaction, while a cognitive state engine tracks hidden dynamics like confidence, persistence, frustration, aggression, social pressure, risk level, user control, and disengagement. As the user responds, the system shows how different choices affect the interaction in real time through an actively undated 3D graph, state readouts, and explanations.

The goal is not to create fear, but to build situational awareness, boundary-setting skills, and confidence. FEMME helps users understand how social pressure and escalation work, and lets them practice safer, clearer responses in a controlled environment.

The prototype for FEMME is an educational simulator for practicing boundary-setting and recognizing social pressure in unwanted interactions. You pick a scenario (bus stop, parking garage, party, workplace, etc.), and an AI character responds to your choices while a real-time cognitive state engine tracks eight psychological variables—confidence, persistence, frustration, aggression, risk level, your control, social pressure, and disengagement. Over 60 rules map your responses to state changes, so saying "no" directly tanks their confidence while apologizing increases it. A 3D graph visualizes the live state, and an explainability panel breaks down what each response triggered. After the interaction, a debrief analyzes your trajectory, highlights your best and worst moves, and gives takeaways. The guardrailing is deliberately strict—no threats or graphic content. I believe it's useful if you want to understand the hidden dynamics of coercion, though it's explicitly not a substitute for real safety training.