From topic to Problem

• The following is mostly from The Craft of Research
• They state that you could consider a project to have four steps
  1. Find a topic
     • You should be able to master a reasonable amount of information in the time allowed.
  2. Question that topic until you find questions that catch your interest.
  3. Determine the kinds of evidence your readers will expect you to offer to support the answer to your question.
  4. Determine if you have sufficient evidence to to provide this support.
• In our case, we might not be looking for a question.
  • Most people who write about research are deeply enamored with a question.
  • I have always found this to be a stumbling point.
  • I guess my question has always been
    • Can I improve or extend knowledge in this area?
    • Can this be done?
    • Can I automate this?
  • But that is probably more my style than the accepted practice.
• Bizup et al point out that you should not focus on the big questions:
  • Is P=NP?
  • Can 42 be expressed in the form $a^3+b^3+c^3=42$ ?
• Focus on something that catches your attention
  • Is it possible to score horseshoes with a ANN?
  • Can I build an AI that better mimics human social networks in a game?
  • How can I make the syntax error output of g++ more manageable for a person with limited or no sight?
• They say, if you feel an itch, start scratching.
• In my opinion, if you read enough about a topic, you will find your question.
• Expertise
  • They point out that you are in the process of becoming an expert.
  • That is what all the reading is about.
• From a broad topic to a focused one.
  • "... your biggest risk is settling on a topic so broad that it could be a subheading in a library catalog."
  • Just keep narrowing.
  • Security in Human Computer Interaction -> Security in accessing web sites -> Secure methods for logging into a web site -> A survey of best practices for implementing user authentication to secure web sites.
  • Notice that the title keeps getting longer and more specific.
  • Look at most paper titles, they frequently are extremely specific
  • Vis Week program for 2019
    • "Modeling and Visualization of Economy Data with Incremental Domain Knowledge"
    • "Visual Analysis of the Time Management of Learning Multiple Courses in Online Learning Environment"
    • "The Effect of Edge Bundling and Seriation on Sensemaking of Biclusters in Bipartite Graphs"
  • Or perhaps look at the KDD 2019 proceedings.
• This is probably not a linear process
  • You may refine too tightly, then need to back off.
  • You may expand to be too wide, and need to refocus.
  • As you explore, you will find questions, or ideas, or whatever you want to call it. Keep track of them!
• They caution that just compiling facts is insufficient
  • You are not writing an encyclopedia entry.
  • You are attempting to find something you can explore further.
  • They return to a question, I return to where can you have an impact?
• They suggest answering "So What?"
  • Is there a reason to do this?
  • If you do this will someone care?
  • If you don't do this, will it matter?
• A major source of inspiration for me is always in the future directions/conclusions sections of papers.
  • They suggest where they will head next.
    • Does this make sense? Do you have a solution?
  • Look at the conclusion to Biship: An Application of a Fast Data Encryption Standard Implementation
• Also, consider
  • Did they make any simplifying assumptions you can remove?
  • Should this be followed by an empirical study?
  • Can this be applied elsewhere?
• In general, read with an open questioning mind.