PyData Manchester

Empathy Data in LLM's - Part II - Lucy Stafford-Hughes

Building on her previous talk in Sept 25 on Empathy Data in LLM's, this talk explores the development of stable interaction patterns in large language models using anchoring and identity containers as a method looking at reducing hallucinations and improving response context and accuracy.

Lucy will look at observations across 2 different AI models on building stable patterns, including interaction with the latest research such as Anthropics work on emotion concepts and emotion vectors in Claude. The talk will also touch on why this matters for human-AI collaboration, and implications across future emergent behaviours.

Lucy is a consultant, writer, and AI researcher. Her areas of expertise include complex business systems, Agile ways of working, and human-AI interaction.

The meaning of lift - understanding marketing effectiveness - Josh Hayes (He/Him)

Data science is about more than just answering questions, it’s about understanding them. Often multiple stakeholders will ask the same, reasonable-sounding question, only for it to become clear that they actually all want something subtly different. The role of a data scientist in this case is to listen to the questions, understand what is actually being asked, and help the stakeholders to develop language to better communicate what they want to measure.

In this talk, Josh will talk through an example of this in marketing, discussing advertising spend, attribution modelling, and how we can measure and talk about the impact of campaigns.

Josh Hayes is a Senior Data Scientist at Autotrader, with nearly 10 years of experience across academia and industry. He holds a PhD in astrophysics, and has led teams of data scientists, analysts, and engineers at Natwest and Autotrader. Among various projects, he has led work calculating carbon emissions associated with financial lending, modelled evolution of the UK’s housing stock, and worked to improve real-time monitoring of key business metrics.

In addition to his professional work, Josh has a long history in public speaking to both technical and general audiences. He has featured at the Bluedot festival, worked as a contributor to BBC radio, and was a long-time presenter of the Jodcast astronomy podcast, alongside speaking at multiple academia and industry conferences.

LOCATION
We'll be at Autotrader, who are also kindly supplying catering. The capacity is limited to 50.
After the talks we'll all head somewhere local for some post-event socialising.

EVENT GUIDELINES
PyDataMCR is a strictly professional event, as such professional behaviour is expected.
PyDataMCR is a chapter of PyData, an educational program of NumFOCUS and thus abides by the NumFOCUS Code of Conduct: see https://pydata.org/code-of-conduct.html
Please take a moment to familiarise yourself with its contents.

ACCESSIBILITY
Toilets and venue are accessible.

SPONSORS
Thank you to NUMFocus for sponsoring Meetup and further support.
Thank you to Autotrader, Krakenflex and Horsefly Analytics for their ongoing support and sponsorship of PyDataMCR.

ACQUIRE YOUR TICKET HERE

Now in its second edition, PyData & PyCon Yerevan 2026 brings PyData and PyCon together to gather the local and global data, research, and Python community in Yerevan.
Led by the American University of Armenia and the AUA Akian College of Science and Engineering, the conference creates a shared space for learning, exchange, and collaboration.
Days full of talks, discussions, and people from different backgrounds in one place.

👉 𝗗𝗼𝗻’𝘁 𝗺𝗶𝘀𝘀 𝗮𝗻𝘆𝘁𝗵𝗶𝗻𝗴 𝗮𝗻𝗱 𝘀𝘂𝗯𝘀𝗰𝗿𝗶𝗯𝗲 𝗳𝗼𝗿 𝘂𝗽𝗱𝗮𝘁𝗲𝘀 𝗮𝘁 https://pycon.am/
𝘐𝘯 𝘤𝘢𝘴𝘦 𝘰𝘧 𝘲𝘶𝘦𝘴𝘵𝘪𝘰𝘯𝘴: 𝘺𝘦𝘳𝘦𝘷𝘢𝘯@𝘱𝘺𝘥𝘢𝘵𝘢.𝘰𝘳𝘨

https://www.meetup.com/pydata-yerevan/events/314378875/

#pydatayerevan #pyconyerevan #pydatapycon #yerevan2026 #PyDataPyConYerevan2026 #techconference