The qualifying exams are a trying event. A candidate must prepare for a wide technical field while also dealing with the personalities of individual committees. This note aims to provide a little guidance on the types of questions committees are likely to ask. It is focused on the breadth requirement of the UIUC Systems and Networking qualifier in CS, which uses a reading list of 25 papers as a foundation. Six types of questions cover most cases. Here are some suggestions on what to do and what not to do when you get them and how to prepare for each type of question.
How to answer. Ideally you know the answer to the question and do exactly what is asked. However, it is just about as important what you do when you do not know the answer as when you do. For example, you may respond by saying I don’t recall the exact steps but I can tell you its purpose followed by an accurate description of the main contribution of the algorithm. Sometimes you will need to say I don’t know. This is best not done too often, but is better than some alternatives, especially when it is the actual state of affairs.
How to answer. When the committee asks a soft ball question it is sometimes good news (we are satisfied you’ll pass so we don’t need to ask you any more hard ball questions) and sometimes bad news (we haven't found anything you are able to do well enough so we’ll give you this one last chance). It is a good idea to prepare some of the papers you especially like in particular detail so you have a chance to shine if you get the chance. However, be prepared to handle some hard ball questions about the paper you have selected as your favorite.
How not to answer. I didn’t like any of the papers, they all bored me or What papers do you mean?
3. Elevator Pitch Questions
How to answer. It is important to prepare each assigned paper with respect to this type of question. Often you can find the most essential pitch in just the abstract of the paper, but just reading the abstracts will not allow you to give a good elevator pitch in your own words. On the other hand, it is easy to read a whole paper, understand each section, and still fail to see the forest because of all the trees. An elevator pitch works much better if you can provide some context that goes beyond the boundaries of the one paper you are discussing. However, keep it short since elevator rides don’t take more than a few minutes, but be prepared to go into more details if asked.
How not to answer. Do not plunge headlong into the details of the paper in your response. Also, do not respond by offering just your opinion of the paper: this may well be something you include, but it is not the first priority in your response.
These are questions intended to determine whether you are familiar with topics that are fundamental to understanding the assigned readings or your area subject generally. For example, What is FEC? or What is the difference between a role and a group?
How to answer. In general you should know the answers to basic questions. Each time you read one of the assigned papers, keep an eye out for the basic topics the paper assumes you understand. If you do not understand these topics, then take the time to learn them at least well enough to follow the paper. When you answer the question you should try to avoid long answers or digressions; the committee members will be especially interested in your ability to answer these questions even if your knowledge is incomplete. For instance, if the committee asks you the fourteenth digit in pi you might say that you don't know off hand, and then show a technique that can be used to calculate the fourteenth digit. Most basic questions you get from committees are not this far off topic, but you can't always count on it. Keep your composure when you respond; nobody knows everything.
How not to answer. Basic questions can throw anyone off guard. If you really don't know the answer it may be best to admit this and hope the committee moves on to something you can answer and doesn’t over-react to a gap in your knowledge. Committees do have the option of making a conditional pass with a recommendation for a course you should take to fill in the basics.
5. Unanswerable Questions
These are questions aimed at testing your knowledge of fundamental unresolved issues. For example, Which is better, packet based networking or circuit based networking? or Is it better to have a centralized solution or a decentralized solution? or Is static checking better than dynamic checking?
How to answer. Such questions typically demand an ability to talk about fundamental tradeoffs in design or analysis and it is essential to draw out one or more of the issues in question. It is very helpful to have examples on hand to illustrate the tradeoffs. For instance, a property that is amenable to static analysis can be contrasted with a property better left for dynamic checking. It is hard to know these types of questions in advance since they are not usually the topic of a single paper but instead are themes that appear across many different studies. An exercise is to think, for each assigned paper, what are the deeper issues that this paper touches upon beyond the more narrowly described objectives in its abstract.
How not to answer. It is important to recognize these types of questions and avoid simplistic answers. For example, Everyone agrees that packet networking is better than circuits. If you have an opinion it may be fine to express it, however. Committees are often looking for candidates with their own ideas and inspiration. For instance, "I think that having decentralized credentials would provide a more scalable authorization system than reliance on a centralized access control list". But be prepared to discuss this fairly with a committee member who does not agree.
6. Stupid Questions.
These are questions intended to test a candidates ability to deal with audiences that are not deeply familiar with topic being discussed. For example, You,ve discussed using forward error correction, but can this technique also correct backward errors?
How to answer. These types of questions often arise in interview talks or with diverse types of audiences. Often the question is more unexpected than stupid. A reasonable strategy is to ask the questioner for more details about their question in hope that a meaningful question can be derived. Occasionally it is reasonable to answer a related question you think is what the questioner meant to ask. Occasionally it is reasonable to improve the understanding of the questioner by repeating an earlier point or giving more details.
How not to answer. It will be tempting to respond to these questions with remarks like Can we put this off until later? or Haven't you been listening to anything I’ve said? but committee members will tend to dislike such responses. Your biggest risk is misinterpreting a clever or provocative question as a stupid one. For example, "What good are public keys if certificates require online validity checking?" If a question really does seem to be meaningless, be respectful in your response but avoid spending lots of time trying to appear to answer. In general you can't tell whether a committee member is trying to test your ability to respond to a stupid question or just asking a stupid question of their own!