- Background
- Purpose
- Problem
- Limitations
- (Method)
- (Outline)
Here follows a description for each level in the Funnel Model:
Level: Top level: You have to start at some suitable high-enough level. A classic beginner's mistake is to start too low down the hierarchy. This is OK in a research paper where all readers are already deep into the subject, but not for a master-thesis report. But do not start at a too high and uninteresting level. Start at the first suitable non-trivial level, and go down the hierarchy very quickly. It is typical to write 1-2 sentences or perhaps a short paragraph. For each class, the best is if you find a book or survey paper as reference. Else, you try to cite one or a few of the main papers on the topic, or at least one paper, for instance a recent one, on the topic.
Level: Which are relevant to you and why? Go down the most relevant branch(es) by explaining their relevance to your work. Again, typically just a sentence or clause per branch.
Level: Which specific alternatives exist in the remaining classes. Again, you narrow down just like for the Top Level (see above).
Level: Which are most relevant to you and why? Again, you continue down the most relevant branch(es) by explaining their relevance to your work, by typically just a sentence or clause per branch.
Level: Describe the individual works' pros and cons w.r.t algorithm/method/study. Finally, you discuss the individual previous scientific papers, and its advantages and shortcomings (if any). I.e., why is your problem not already solved in that paper for your research question. It is often enough with just 1-2 sentences or clauses for each paper.
This hierarchical structure also automatically provides you with the substructure of your Previous Work, e.g., the names for your subsections of previous-work classes.
It is rather common that students claim that there does not exist suitable related work. I can understand the feeling, but there always exist the most related work even if that work would happen to be rather far away (however, that is unlikely and has never happened in the ~150 thesis works that we have supervised). The problem is often to find the first relevant paper. Use search phrases from your topic and research question to google and find a few starting papers. Then use the following algorithm to find the relevant previous work:
1. Find_Relevant_Previous_Work( Paper P ):
2. Use all relevant references, R1…Rn, in paper P.
Find_Relevant_Previous_Work( Rj ). // I.e., recurse via the paper's references (to find earlier relevant research papers).
3. Also search on, e.g., Google Scholar for all papers, C1…Cn, that cites this paper P.
Find_Relevant_Previous_Work( Cj ). // i.e., recurse for the paper's citations (to find later research papers).
Terminate recursion when it no longer leads to new relevant references. For each found relevant paper, make a note of its:
• relevance to your study
• relevant advantages and shortcomings. I.e., why is your problem not already fully solved by this paper.
Now, you have everything to write down your Previous-Work section. You can find an example of Previous Work with the Funnel model here.
Obligatoriskt enligt högskoleverket. Kort stycke om etiska aspekter.
Example 1:
"Ethical Considerations
After careful consideration, we have not recognized any societal, ethical, or ecological
concerns that directly apply to this work, its results, nor its potential future. However,
in the broader context, we do recognize that more realistic simulations could
ease the creation of false scenarios for the purpose of deception. Realistic oceans
by themselves might not have any obvious applications for such use, but they could
be used in combination with other realistically simulated and rendered models. Another
concern is that such realism would further entice humans away from the “real
world”, which may negatively affect both physical and mental health. Nevertheless,
such concerns plague the entire field of computer graphics, in contrast to the small
scope addressed in this report. Moreover, it is important to consider these concerns
in the light of all the positive outcomes that realistic computer graphics bring – from
medical research and auto-motives, to architecture and entertainment.”
Example 2:
"The ethical considerations for this thesis are minimal. As the work will be based on
data fabricated and collected by the authors, no breach of privacy or other sensitive
information can be realized. Moreover, the practice of high-performance computing
consumes large amounts of energy with the powerful hardware utilized. Although
this project aims to improve energy efficiency of GPU computations, energy consumption
for calculations is something to consider in general from an environmental
point of view.
Example 3:
Ethical considerations
With advancements in computer graphics, the difference between reality and for
example a game becomes smaller. This could have detrimental effects on issues
such as game addiction where people spend an unhealthy amount of time in front
of a computer screen. Too much time in front of a computer screen can be bad for
peoples eyes as well as body with increased risk of blood clots through sitting down
for long periods of time.
More realistic lighting and graphics when it comes to Virtual Reality could lead
to people neglecting real life to spend most of their time in a virtual world instead.
On the other hand, the immersion that comes with more realistic graphics in games
can lead to them being more enjoyable to play, as long as the playing time is kept
at reasonable levels.
Example 4:
Ethical Aspects
In this thesis, there are both positive and potentially negative aspects to discuss
regarding ethics and sustainability.
On one hand, this project could be used to analyse sustainability in the environment,
by performing different experiments like mentioned above. As shown by these
experiments, this project could for instance:
• Help improve decision making regarding which infrastructure expansions are
reasonable and could be ethically motivated with concern to wildlife.
• Help determine what extent of hunting should be done for certain species.
This project could also act as another motivation to combat climate change when
seeing how it affects certain ecosystems and animal populations. In this thesis, the
consequence of sea level increases was observed, but other consequences could also
be analysed, such as how the animals’ food/water sources are affected, how their
breeding is affected, etc.
On the other hand, there are also potential risks to consider. As with all projects
that involve some sort of artificial intelligence, there are always malicious use cases.
In the context of this thesis, one such potential use case could be that the agents
that were trained in a specific environment, might be useful in some kind of military
sense, which could be dangerous if they received more accurate input for their
perceptions.
However, as the project stands now, this is considered extremely unlikely, and the
use cases are more limited to the analysis of ecosystems than anything else.
Your references (along with Previous Work), are arguably some of the most important parts of your thesis. It is by having correct references and putting your work in relation to all the most relevant references that you show the reader that you have solid knowledge of your field. Additionally, you clearly show how your work complements and expands previous knowledge. The reader can value your contributions, based on what others have done before.
However, more importantly… in your thesis you make some kind of experiment or investigation, which is based on previous scientific learnings, e.g., algorithms, maths, etc. You assume that those learnings, maths, or algorithms are correct, and you can (in principle) check that by reading those papers. Those papers also start from assumptions, which you recursively can check via their references. The papers also typically add scientific learnings via conclusions from their experiments, and these are typically repeatable so you could in principle redo them if you don’t trust them.
In this way, you can think of your thesis as a brick stone that you add to the top of a brick wall of scientific knowledge. Your references are the glue for the reader to know that your brick stone (your contributions) stand solid. Your experimental results are your new scientific contributions. So, it is essentially your references that makes your text scientific.
In principle, you should have a reference after every (non-trivial) claim you make in your text. This could be after almost every sentence, for some of your paragraphs. However, if several sentences contain claims backed by the same references, you can state the reference once in the paragraph (in computer graphics, this is often where you make the first such claim, and in other research fields, it might be at the end of the paragraph).
Sometimes or even often, it becomes ugly to have the same references repeated over and over again, for paragraph after paragraph or section after section. Then, you may remove repetitious ones so your text visually looks better (so it does not look stupid :-) ).
You can take a look at some of our previous master-thesis reports here. You find a very good example here.
Many of the references come in the Previous Work section. However, even in the rest of your text (except Abstract and Summary) you should also use references.
For the references, there are different systems you can use. You can for instance cite using [1], [2]... or [Kim 1998], [Kim and Andersson 1998] or [Kim et al. 1998] (if there are more than three authors).
References should be listed with the following information:
Book: (at least) author, book title, publisher, year.
Book chapter: author, name of chapter, book title, publisher, pages xx--yy, year.
Journal: author, title of article, title of journal, issue, number, pages, year.
Article in proceeding (typically most common): author, title of article, name of proceeding (e.g. Proceeding of SIGGRAPH 2002), pages, (month), year.
I.e., in general, the following information is required: author, title of text, name of publication place (including issue and number for journals), pages, (month), year.