Computer Laboratory

Course pages 2017–18

Interaction with machine learning

Principal lecturer: Prof Alan Blackwell
Additional lecturer: Dr Advait Sarkar
Taken by: MPhil ACS, Part III
Code: R230
Hours: 16 (8 2-hour sessions)
Class limit: 16 students
Prerequisites: Students will benefit from pre-requisite content in machine learning that is delivered in Michaelmas.

Teaching Style

This is an advanced course in human-computer interaction, with a specialist focus on intelligent user interfaces and interaction with machine-learning and artificial intelligence technologies. The format will be largely *Practical*, with students carrying out an original empirical research investigation over the course of one term. All empirical studies will address human interaction with some kind of model-based system for planning, decision, automation etc. Possible study formats might include: System evaluation, Field observation, Hypothesis testing experiment, Design intervention, Corpus analysis, or others as shown to be appropriate from evidence of prior research publications that have adopted specific empirical formats.

Aims and Objectives

The goal of this course is to develop professional research skills in the application of machine learning methods as a component of interactive system design. Students should be able to demonstrate competence in producing a specialist research publication in a field such as intelligent user interfaces (eg ACM IUI), or interaction with intelligent systems (eg ACM TIIS).

Taught content will address contemporary issues in interaction with machine learning systems, including interactive labelling, visualisation, explanation, program synthesis, attribution of intention and the ethics of artificial intelligence. Weekly sessions will consist of one hour of taught content (often delivered by guest lecturers, drawing on contacts at the Centre for the Future of Intelligence, the Alan Turing Institute, Microsoft Research and others), followed by an hour in which students present progress reports and discuss methodological questions related to their empirical projects. This will be a total of 16 hours contact time.

Assessment

There will be a minor assessment component (20%) in which students compile a reflective diary throughout the term, reporting on the weekly sessions. Diary entries should include citations to any key references, notes of possible further reading, summary of key points, questions relevant to the personal project, and points of interest noted in relation to the work of other students.

The major assessment component (80%) will involve a report on research findings, in the style of a submission to the ACM CHI or IUI conferences. This work will be submitted incrementally through the term, in order that feedback can be provided before final assessment of the full report. Phased submissions will cover the following aspects of the empirical study:

  1. Research question
  2. Method
  3. Literature Review
  4. Introduction
  5. Results
  6. Discussion / Conclusion

Feedback on each phased submission will include an indicative mark for guidance, but with the understanding that the final grade will be based on the final delivered report, and that this may go up (or possibly down), depending on how well the student responds to earlier feedback.

Reflective diary entries will also be assessed, but graded at a relatively coarse granularity corresponding to the ACS grading bands, and with minimal written feedback. Informal and generic feedback will be offered verbally in class, and also potentially supplemented with peer assessment.