What is done in a workshop? How does it differ from an academic conference?

The reason why it is difficult to tell what workshops are about is because it is a catch-all category that many different types of academic meeting are labelled as. To illustrate, let me give examples of the nature of some of the events that I have attended in the last two years that all use the same word "workshop" to describe themselves:

• A "baby conference" attached to a full-size conference, where the small meeting simply isn't large enough to meet on its own yet.
• A project meeting for researchers who are all funded by the same large grant
• A planning and discussion session aimed at helping determine the direction of a field
• A joint industry/academia fact-finding meeting sponsored by an industry consortium
• A specialty conference attended by around 100 people
• A premier conference attended by several hundred people
• A working meeting by a standards development group

The length of these meetings ranged from a single afternoon to a full week. Their programs ranged from nothing but loosely structured discussion to a full-on tightly packed conference schedule. The level of peer review ranged from non-applicable to minimal to full-on single-blind review and revision.

In short: a workshop is whatever it wants to be, and different ones serve different purposes in the academic ecosystem.

In CS subfields related to HCI and related topics of communication between users and software, a workshop is almost the same as a conference, that is, there are speakers who give talks about their papers. However, the following differences can be observed:

Organisational:

• A workshop lasts just one or half a day.
• A workshop is "embedded" into a conference; it doesn't take place on its own, but it takes place at the day before or after the main conference. Sometimes, conference registration includes workshop registration.
• The audience tends to be much smaller. This is partially because often, several workshops are scheduled at the same time, and partially, because not all conference guests attend any workshops (especially when the workshop needs to be registered and paid for separately). In any case, it is not untypical for smaller workshops to have just a handful of attendees, each of which will present something.

Contents:

• There is less of an expectation of presenting fully finished work. Work in progress, or results based on a preliminary study, are usually accepted, if not explicitly invited.
• Likewise, remarks about future work can be a bit more central than in normal talks, as the workshop may prove as an opportunity to find collaborators who would like to help tackle the suggestions for future work.
• While conference sessions typically feature an opportunity for questions and answers on every single paper right after each talk, workshops sometimes schedule an additional (sometimes considerable in duration) wrap-up discussion that is supposed to provide some time for discussing everything that was presented during the workshop, identifying common issues and chances, and possibly developing ideas for further work together.

Sometimes, a workshop is just a conference, but the organizers decided to call it a workshop instead for one reason or another.

So while a lot of people will have some idea of what the difference should be, in practice the difference can be non-existent, so one will need to look at the details of each individual conference or workshop to decide which will be of interest.

(Note that my entire experience is from mathematics, so it might be different in other fields).