This section explains what you have to do before actually conducting a study or experiment.
What question do I want to answer?¶
Write a study script containing:
- What is the hypothesis?
- What conditions exist?
- How was the topic analyzed in related work?
- Where will the results be published?
- Which results will be published?
- Related work on the topic or planned study?
- How did related work collect data?
What are my measurements (what measurements do I need to support my hypothesis)?¶
- Task performance?
- Mutual gaze?
- Length of interaction?
- Facial expression?
What do I want to use in order to answer the question?¶
How will recorded data be analysed in order to support hypothesis? In detail:
- What study do I want to conduct: Specify study format
- Live interaction with an autonomous system
- Wizard-of-Oz style interaction
- Experiment at a local PC
- Online questionnaire (e.g. with videos)
- Which demonstrator(s) and/or hardware do I need?
- Which room/place can I use (check for noise/availability/equipment)
- What time do I need in total (incl. buffer)?
Define Milestones and contingency plans include time for pre-tests to
- test if the whole procedure works as planned/imagined (e.g. where should experimenter be at what time, when does s/he leave the room, when will the participant start, how long does the experiment take...)
- test if the recorded data has the expected quality and contains the information that is wanted
- test if the tasks that the participants are required to perform have the right level of difficulty in order to achieve a good amount of variance (not too easy, not too difficult) test how long it takes to do one run
Setup buffer-time between subexperiments Study duration
- how long are you going to conduct your study?
- will you need longer than 6 month?
Think about charging time of devices. E.g. robot base.
What needs to be recorded?¶
Specify exactly what data or modalities you need to have recorded. Sometimes, it is better to concentrate on only the things you need to answer your hypothesis. Sometimes, other people (from your group) may also be interested in the data. So talk to them first.
How many participants per cell are needed? Which participants should participate (age, gender, experiences in robotics etc)?
- is an ethical votum (Ethikantrag) necessary? If yes: + How to determine? + Whom to ask? + Procedure?
- what rights on the data should I get from the participants
- Naming of participant ids and trials consistent over all parts of the study
- TODO: Exemplary scheme
- how many participants do I need?
- what is the expected power of my experiment https://en.wikipedia.org/wiki/Statistical_power
- do not acquire participants before testing your setup
- use doodle or foodle, better https://www.dfn.de/dienstleistungen/odfnterminplaner/
- anonymous doodle so participants can’t see each other
- collect contact information of participants
- only one participant per time slot
- more diverse is better. Not only students
- create appealing flyer
- what information do participants need to know e.g. how to not bias them
- if you have different condition randomize participant assignment