Human-in-the-Loop Machine Learning

6 EC

Semester 1, period 2

5204HITL6Y

Owner Master Artificial Intelligence
Coordinator Nanne van Noord
Part of Master Artificial Intelligence,

Course manual 2024/2025

Course content

The ultimate goal of machine learning (ML) and AI is often considered to be that of building fully autonomous systems. However, every stage of building these systems involves humans, from the data, to design, and deployment. Understanding this interaction with humans, as well as the influence (both intentionally and unintentionally) that humans may have on AI systems and AI systems on humans is at the core of this course. Questions surrounding how to design and build AI systems which prioritise and adapt to human needs and preferences, as well as considerations on whose values are being encoded and what trade-offs are being made will be discussed.

This course is centred around the role of humans in all stages of ML and AI systems, including discussion of present techniques for integrating human intelligence and methods for designing systems centred on humans. Topics addressed in the course include: (inter)active learning, crowdsourcing, machine teaching, learning from human feedback, and human-centred evaluation.

Study materials

Literature

  • Reading list of academic articles will be provided.

Objectives

  • The student can understand key concepts in human-centred data work (e.g. evaluation, annotation, crowdsourcing).
  • The student can understand techniques for building systems that adapt to human feedback ((inter)active learning, RLHF, machine teaching).
  • The student can analyse a given intelligent system to identify sub-components that can be improved for humans.
  • The student can evaluate the performance of a hybrid human-AI system and identify sub-components in need of improvement.
  • The student can create a human-centred AI system that effectively solves a problem of practical consequence.

Teaching methods

  • Lecture
  • Presentation/symposium
  • Self-study
  • Working independently on e.g. a project or thesis

Learning activities

Activity

Hours

Deeltoets

2

Hoorcollege

24

Presentatie

2

Werkcollege

24

Self study

116

Total

168

(6 EC x 28 uur)

Attendance

This programme does not have requirements concerning attendance (OER part B).

Assessment

Item and weight Details

Final grade

0.3 (30%)

Assignments

1 (50%)

Presentation - “Values”

1 (50%)

Report - “Evaluation”

0.3 (30%)

Course Project

Must be ≥ 5.5, Mandatory

0.4 (40%)

Exam

Must be ≥ 5.5, NAP if missing

Fraud and plagiarism

The 'Regulations governing fraud and plagiarism for UvA students' applies to this course. This will be monitored carefully. Upon suspicion of fraud or plagiarism the Examinations Board of the programme will be informed. For the 'Regulations governing fraud and plagiarism for UvA students' see: www.student.uva.nl

Course structure

Weeknummer Onderwerpen
1 Introduction + ML for social good
2 Values in ML + Benchmarking and Beyond
3 Human-Centred Evaluation + Machine Teaching
4 Crowdsourcing + Values in data
5 Interactive Learning + Human Feedback
6 Active learning + Human-Centred Explainability
7 Project week
8 Exam

Contact information

Coordinator

  • Nanne van Noord