Summary and Setup

Workshop overview

Machine learning and AI are now routine tools in astronomy. They are used to classify objects, generate catalogues, estimate physical parameters, and prioritise observations across increasingly large and complex datasets.

This workshop focuses on how to use these tools responsibly, reproducibly, and transparently in a research context. Rather than teaching ML techniques themselves, the emphasis is on how ML fits into the broader scientific workflow: how results are produced, documented, reused, and interpreted over time.

The intent of the workshop is practical. It aims to help participants recognise common failure modes in ML‑based astronomy, understand how reproducibility and FAIR practices apply to data and models, and adopt lightweight habits that make their work easier to validate, defend, and reuse. Ethical issues are discussed in terms of scientific responsibility, scope control, and downstream reuse, rather than policy or regulation.

This workshop is designed to be discussion‑driven and reflective. Participants are encouraged to think about their own projects and workflows, rather than aiming for idealised or perfect practice.

Assumed background and skills

This workshop assumes that participants:

  • are familiar with basic astronomical data analysis
  • have some exposure to machine learning or AI methods, either through use or collaboration
  • have experience working with real survey or simulation data

We assume no prior knowledge of:

  • formal research ethics frameworks
  • FAIR data principles
  • software engineering best practices

The workshop does not teach:

  • how to build ML models
  • specific ML algorithms or libraries
  • advanced software tooling

Instead, it focuses on the decisions and documentation that sit around ML use in astronomy research.

Who this workshop is for

This material is particularly aimed at:

  • PhD students using ML or AI in their research
  • early‑career researchers working with ML‑derived results
  • astronomers collaborating with data scientists or ML specialists
  • researchers preparing ML‑based results for publication or reuse

The examples and activities are drawn from astronomy research contexts and are intended to be directly applicable to real projects.

What you should expect to take away

By the end of the workshop, participants should be able to:

  • articulate what reproducibility means in ML‑based astronomy
  • recognise where ML workflows commonly break reproducibility and FAIR practices
  • identify ethical risks related to scope, reuse, and over‑interpretation
  • apply a small set of practical actions to improve their own workflows

You should leave with a clearer sense of what is reasonable to do now, and what can be improved incrementally over time.

This workshop is not about doing everything right. It is about making ML‑based astronomy easier to trust, reuse, and understand.

There will be no coding or computer based exercises used in this workshop. You need only have a browser to view the content, and maybe a notepad to take some notes.