Half of 15-year-old girls in England lack the confidence to speak in public. The neuroscience tells us why — and exactly what to do about it. This is the story of how an enterprise AI interview tool became a speech-training experience for the next generation of Bad Girls.
The research is unambiguous. The window for intervention is narrow. The tools to act have never been more accessible.
Adolescence · Ages 11–14
Harvard psychologist Carol Gilligan documented what she called the "voice crisis" in adolescent girls — a phenomenon where girls who were outspoken and confident at age 10 become hesitant, self-censoring, and publicly silent by age 14. This is not a personality trait. It is a learned behaviour, driven by social conditioning, fear of judgment, and the absence of spaces where girls are invited to speak with authority.
The neuroscience explains the mechanism. During adolescence, the amygdala — the brain's threat-detection centre — becomes hyperreactive to social evaluation. Simultaneously, the prefrontal cortex, which regulates emotional responses and supports confident self-expression, is still developing. The result: speaking up in public feels genuinely dangerous to a teenage girl's brain. Without deliberate intervention, this threat response becomes the default.
Intervention Research
Speakers Trust's impact research on its flagship Speak Out Challenge programme — which operates in over 500 state secondary schools across England — demonstrates that structured speech practice leads to measurable increases in skills, confidence, and willingness to speak publicly. The intervention works. The question has never been whether speech training helps. The question is how to make it accessible, repeatable, and — critically — something a young person actually wants to do.
The gap is not in the evidence. It is in the delivery mechanism. Schools cannot provide daily, personalised speech practice at scale. We can.
The design of Speak Like a Bad Girl was directly inspired by a category of enterprise AI tools built to help adults prepare for high-stakes conversations. We asked: what if this existed for children?
Enterprise Speech Coaching · Direct Inspiration
Asgard AI is an enterprise-grade agentic AI platform that enables organisations to build structured AI workflows — including speech-to-text input pipelines, multi-turn conversational agents, and structured response evaluation systems. Its architecture demonstrates a critical insight: that speech can be captured, transcribed, and evaluated against structured rubrics at scale, without human evaluators in the loop.
The broader category of AI interview preparation tools — including platforms like Yoodli, which uses speech analysis to coach professionals on clarity, filler words, pacing, and confidence — has proven that AI can deliver meaningful, personalised speech feedback. These tools are used by adults preparing for job interviews, sales pitches, and board presentations. They work because they provide immediate, specific, non-judgmental feedback in a private environment.
The question we asked was simple: why does this category exist only for adults preparing for corporate interviews? The stakes of learning to speak confidently are highest in childhood and adolescence, when the neural pathways for self-expression are still being formed. If a 35-year-old professional benefits from AI speech coaching before a job interview, a 12-year-old girl benefits from it before every classroom, every friendship, every moment she decides whether to speak or stay silent.
Explore Asgard AIThe brain is not fixed. Repeated, structured speech practice produces measurable changes in the neural circuits that govern confidence, self-regulation, and emotional expression.
Core Mechanism
When a young person is asked to speak in public, the amygdala — the brain's threat-detection system — fires a fear response. This is not metaphorical: neuroimaging studies show elevated amygdala activation during public speaking that is comparable to responses to physical threat. In adolescents, this response is amplified because the prefrontal cortex (PFC), which regulates the amygdala and supports deliberate, confident self-expression, is still maturing.
The intervention is neuroplasticity. Repeated exposure to speech — in a low-stakes, structured, non-judgmental environment — gradually strengthens the prefrontal-amygdala regulatory pathway. The brain learns, through repetition, that speaking is not dangerous. The threat response diminishes. The capacity for confident self-expression grows. This is not motivational language. It is the documented mechanism of speech therapy, cognitive-behavioural intervention, and deliberate practice research.
Research on training-induced neural plasticity in children and adolescents (PMC, 2021) demonstrates that structured skill training produces measurable changes in brain structure and function. The same mechanism that allows a child to learn a musical instrument — repeated, effortful practice producing lasting neural change — applies to speech. Each time a young person practises speaking under structured conditions, they are not just improving a skill. They are physically reshaping the neural circuits that govern self-expression.
Practice is not preparation for confidence. Practice is the mechanism of confidence.
The prefrontal cortex exerts top-down regulatory control over the amygdala. When this pathway is strong, emotional threat responses — including the fear of speaking — are modulated before they become paralysing. Research on non-invasive neuromodulation (Kelley et al., 2019, PMC) shows that stimulating prefrontal activity promotes successful self-regulation by altering the balance between the PFC and subcortical threat systems. Deliberate speech practice achieves the same effect through behavioural means: it trains the PFC to engage, and the amygdala to stand down.
Every time a girl speaks when she is afraid, she is training her prefrontal cortex to be stronger than her fear.
A 2022 review in the Journal of Voice (Barbosa et al.) on voice and self-regulation found that voice use is directly linked to emotional regulation capacity, particularly in adolescence. The act of speaking — not just thinking, but vocalising — activates self-regulatory neural networks. A 2024 study in PMC on neural effects of one's own voice found that listening to one's own voice activates unique neural processing pathways distinct from hearing others, supporting self-monitoring and emotional regulation. The voice is not just a communication tool. It is a neuromodulation tool.
Teaching a girl to use her voice is teaching her to regulate her own nervous system.
The prefrontal cortex continues developing until the mid-20s, with the most rapid development occurring between ages 10 and 16. This is both the period of greatest vulnerability — when the amygdala-PFC imbalance is most pronounced — and the period of greatest neuroplastic opportunity. Interventions during this window have disproportionate long-term impact. A short CBT-based speech training programme of five sessions was shown to be effective in treating speech anxiety in youth (de Jong et al., 2021). The window is open. The question is whether we use it.
The same neuroplasticity that makes adolescence difficult makes it the most powerful time to intervene.
The feature is not a quiz with a microphone. Every element — the judges, the questions, the four pillars, the verdict format — was chosen to maximise the psychological safety and neurological benefit of the practice.
The scoring is delivered by Gargi, Draupadi, Vishpala, Amrapali, Chitrangada, and Sanghamitra — six women who each faced the exact challenge we are asking the player to face: speaking with authority in a world that told them to be silent. This is not decoration. It is the psychological frame. The player is not being evaluated by a machine. She is being heard by women who understand what it costs to speak. The narrative reframe changes the emotional experience of receiving feedback.
Emotionally safe feedback is more likely to be internalised and acted on than algorithmically delivered scores.
Most speech assessment frameworks score clarity and structure. We added Conviction and Boldness as equal pillars. This is deliberate. The research on girls' voice loss shows that the primary failure mode is not unclear speech — it is self-censorship, hedging, and the suppression of strong opinions. By scoring Boldness as a first-class metric, we send a direct message: having a strong point of view is not a social risk. It is a skill. It is valued. It is what the Bad Girls did.
What you measure is what you teach. Scoring boldness teaches girls that boldness is a virtue.
Each Bad Girl asks a question drawn from her own historical context — but reframed for the player's lived experience. Gargi asks about a belief the player holds that others question. Draupadi asks what the player would do if the rules were unfair. These are not comprehension questions. They are invitations to have a point of view, to defend it, and to speak it aloud. The content of the speech matters less than the act of forming and expressing a genuine position.
The question is the intervention. Asking a girl what she believes is the first step to teaching her to say it.
The experience is designed to be used alone, on a phone, without an audience. This is the key design constraint borrowed from enterprise interview tools: the low-stakes private environment is what makes the practice possible. A girl who would never raise her hand in class will speak into her phone. The neurological benefit of the practice is identical — the amygdala-PFC pathway is trained regardless of whether anyone is watching. The private practice builds the capacity for the public performance.
The goal is not to perform for the app. The goal is to build the neural capacity to perform for the world.
The feature is live. The vision is larger: a generation of girls who have practised speaking to Gargi, Draupadi, and Vishpala — and who carry that practice into every room they enter.
A girl aged 10–16 uses Speak Like a Bad Girl on her phone, alone, before school. She practises answering questions from women who were braver than she feels. She gets scored. She tries again. Over weeks, the practice accumulates. The neural pathway strengthens. The classroom feels less threatening. The hand goes up more often.
Speak Like a Bad Girl is designed to complement programmes like Speakers Trust's Speak Out Challenge. Where those programmes provide structured group workshops, this provides the daily private practice that makes the workshops land. The combination — structured group coaching plus AI-assisted private practice — mirrors the most effective models in sports training, music education, and language learning.
We are not building a speech app. We are building a neuromodulation tool disguised as a game. The distinction matters. A speech app teaches technique. A neuromodulation tool changes the underlying neural architecture that makes confident speech possible or impossible. The Bad Girls are not just characters. They are the psychological scaffolding that makes the practice emotionally safe enough to do repeatedly — and repetition is the mechanism of neural change.
The enterprise AI tools that inspired this feature exist because organisations understand that the ability to speak with authority is worth investing in. We are making the same investment in children — specifically in girls, at the exact developmental window when that investment has the greatest neurological return.
Almost half of girls fear public speaking
Speakers Trust, 2023
Asgard AI — Agentic AI Builder Platform
Asgard AI
Yoodli — AI Roleplay Platform for Communication Coaching
Yoodli
Training-Induced Neural Plasticity in Youth: A Systematic Review
PMC / Frontiers in Human Neuroscience, 2021
Stimulating Self-Regulation: Non-invasive Brain Stimulation and the Prefrontal Cortex
Kelley et al., PMC, 2019
Neural Effects of One's Own Voice on Self-Talk for Emotion Regulation
Jo et al., PMC, 2024
Treating Speech Anxiety in Youth: A Randomised Controlled Trial
de Jong et al., Journal of Anxiety Disorders, 2021
Voice and Self-Regulation: Integrating Review of the Literature
Barbosa et al., Journal of Voice, 2022
Gilligan: Girls Lose Voices
The Harvard Crimson, 1991
Six Bad Girls are waiting to hear what you have to say. They have faced kings, philosophers, and armies. They can handle you.