
309: Conversational AI: Trust, Values and the Illusion of Understanding with Dr Malak Sadek
In episode 4 of our AI, Ethics and Trust in Coaching series, host Smaranda Dochia sits down with Dr Malak Sadek, researcher and teaching associate at the University of Cambridge's Centre for Human-Inspired AI, to explore why conversational AI affects us so differently from other technologies. Conversation is already natural to us, so we never have to learn how to "use" a chatbot the way we'd learn a new app, and that ease is exactly what blurs the line between a fluent response and genuine understanding. Malak explains how design choices in tone, voice, and responsiveness shape trust, reinforce gender and racial biases, and lead people to attribute empathy, authority, or even relationship-like qualities to systems that have no lived experience or personal stake in the outcome. The discussion also touches on how repeated interaction with these systems may quietly shape behaviour, beliefs, identity, and the choices people make over time.
The conversation then turns to the values built into these systems by default. Malak makes the case that no version of "neutral" really exists: privacy settings, curiosity, playfulness, politeness, cultural assumptions, and commercial incentives are all design decisions that shape how an AI behaves, and many carry Western-centric assumptions that go unexamined. For her, responsible AI needs more than technical safety — it needs meaningful participation, cultural representation, and transparency about the values embedded in the system, so that design strengthens human agency rather than quietly eroding it. She closes with a challenge to coaches and other professionals: rather than resisting these tools, the more useful path is to understand their limitations well enough to guide clients through them.
You will learn:
· Fluency isn't understanding – coaches need to help clients see this distinction
· No system is truly neutral: Every design choice — tone, defaults, privacy settings, cultural framing — embeds values, often Western-centric ones, and coaches should be aware of whose assumptions are shaping the interaction.
· Rather than avoiding AI tools, coaches and professionals are better placed to guide clients by educating them on AI's limitations.
"We have a very strong tendency to give human traits and human characteristics to things that aren't human beings just because they kind of either look human or they remind us of humans somehow. and obviously with LLMs this is on steroids now because they are actually behaving on the surface like, human beings."
Don't forget to subscribe, rate, and leave us a review! Your feedback helps us bring you more valuable content.
For the episode resources and guest bio, please visit:
Flere episoder fra "Coaching Voices: The Association for Coaching Podcast"



Gå ikke glip af nogen episoder af “Coaching Voices: The Association for Coaching Podcast” - abonnér på podcasten med gratisapp GetPodcast.








