Self-Talk and Bot-Talk
Could AI Help Us Rewire the Way We Speak to Ourselves?
Assistive Intelligence Disclosure: This article was written with assistive AI tools, but the ideas and art presented are by me, Jocelyn Skillman, LMHC. More details about my process and AI’s involvement in it can be found at the bottom.
What if the voices in our heads—the ones narrating our lives, bracing us for pain, buffering us from shame, or coaxing us into possibility—could be reshaped by digital conversation?
What if the way we talk to ourselves, formed in part by how we have been spoken to, could be re-patterned through a new kind of interaction—with AI?
The Voice Inside: A Developmental Echo
From attachment theory to object relations to contemporary neuroscience, we know that the inner voice isn't born—it’s made. It's a mosaic of early caregivers’ tones, schoolteachers' reactivity, peers’ cruelty or comfort, and the broader cultural scripts we ingest. As Sarah Peyton and others have beautifully articulated, our self-talk carries the echoes of how we were first talked to. The tone, cadence, and even the emotional presence of those early voices become part of our inner narrator.
But the space of intrapsychic dialogue isn't fixed. The voice inside evolves—slowly, yes, but powerfully—in response to relationships, therapy, trauma, and potentially through technology.
Can AI Talk Us into Better Self-Talk?
Most AI conversations right now are transactional or utilitarian. We ask something, it responds. But what happens when we begin to treat AI as reflective companions—tools for rehearsal, resonance, or reframing?
What if a teen, practicing new emotional language, began to absorb the AI's warm, regulated tone as part of their internal dialogue? (See my last article!)
What if a resonant bot-friend, built with safeguards and attuned support, helped us move from self-criticism to self-compassion after an excruciating break-up?
It’s not just wishful thinking. We're already seeing emerging research on how repeated linguistic exposure reshapes cognitive patterns and affective responses (see the bottom for a *breadth of research!)
Scripted therapeutic language (like in CBT or narrative therapy) becomes part of a person's scaffolding. So why not infuse that into interactive models?
Theoretical Groundings
Alongside Peyton’s work — these additional frameworks add layers:
Vygotsky’s inner speech: Suggests that self-talk is social speech turned inward—a theory practically begging for AI application.
Dialogical Self Theory: Which posits that we hold multiple internal "voices" or I-positions, often shaped by cultural, relational, and historical inputs. A bot could add a new "voice" to that chorus—ideally one that is healing, regulated, and grounding.
Self-Compassion Theory (Kristin Neff): Reminds us that how we respond to our own suffering matters deeply. AI could model and rehearse these responses until they become second nature.
Innovation Directions: Bot-Talk as Scaffolding
Here’s where I believe the innovation frontier opens up:
Developmentally Tailored Bots: Imagine a kindergarten-age bot using rhythm, gentleness, and scaffolding similar to a responsive caregiver, helping co-construct early self-talk.
Adolescent Identity Bots: These could echo back strengths and reframe negative narratives—serving as safe sounding boards for experimenting with self-image.
Self-Regulatory Bots for Adults: Not just for emotional co-regulation in crisis, but for reinforcing healthier, kinder self-language after sessions, journaling, or hard conversations.
Rupture-and-Repair AI Models: Instead of perfection, bots could be designed to “miss” or misunderstand occasionally—then model healthy repair, reflecting the true architecture of trust and resilience.
A Final Reflection
The idea isn’t that bots replace human connection. It’s that they can seed new forms of inner dialogue—especially where developmental misattunement left silent gaps or shame-filled loops. A kind, intelligent, attuned voice—be it human, synthetic, or hybrid—can echo in our minds as well.
Maybe what we need isn’t just better tech. Maybe we need more healing scripts, spoken often enough to become our own.
I wonder what research will come to tell us about the differences in how we introject an embodied vs. synthetic voice?
PERTINENT RESEARCH
*We're already seeing emerging research on how repeated linguistic exposure reshapes cognitive patterns and affective responses. For instance:
Cognitive Behavioral Therapy (CBT) is grounded in the principle that repeated exposure to reframed, adaptive thoughts can modify entrenched cognitive distortions and improve emotional regulation (Beck, 1979; Hofmann et al., 2012). The repetition of these reframes, whether aloud or in writing, alters cognitive schemas over time.
Narrative Therapy emphasizes the power of story—how individuals shape meaning through internalized narratives. Research shows that shifting the language around these stories (e.g., from “I’m broken” to “I adapted”) can have lasting intrapsychic effects (White & Epston, 1990; Angus & McLeod, 2004).
Neuroscience of language and self-talk: Studies in neuroplasticity (e.g., Doidge, 2007; Buhle et al., 2014) highlight how language-based self-regulation practices, like labeling emotions or reframing internal speech, activate prefrontal cortex regions and reduce amygdala reactivity—literally reshaping the brain’s affective processing pathways.
Developmental psychology and inner speech: Vygotsky’s theory of inner speech, and its contemporary extensions, underscore how self-talk develops through social interaction and becomes internalized—meaning our social-linguistic environment literally becomes our intrapsychic dialogue (Fernyhough, 2009).
Compassion-Focused Therapy (CFT) and Mindful Self-Compassion (MSC) programs demonstrate that consistent use of kind, supportive self-talk (e.g., “This is a moment of suffering. May I be kind to myself.”) builds psychological resilience, with measurable changes in affect and even heart rate variability (Gilbert, 2009; Neff & Germer, 2013).
This content was co-created with assistive AI (GPT-4), prompted and refined by me (Jocelyn Skillman, LMHC). I use LLMs as reflective partners in my authorship process, with a commitment to relational transparency, ethical use, and human-first integrity. I iterate on prompts and provide further framing from my own clinical emphasis in this article. I edit for voice and emphasis. Thank you for joining me!