CSE 331 Mini Project

Fall 2017

Details and motivations for the mini project.

Motivation

CSE 331 is primarily concerned with the technical aspects of algorithms: how to design them and then how to analyze their correctness and runtime. However, algorithms are pervasive in our world and is common place in many aspects of society. The main aim of the mini-project is to have you explore in some depth social implications of algorithms.

Just to give two examples for such implications:

Algorithms are pervasive in financial transactions and these algorithms have consequences beyond just trading:
Big data is hot these days and there is a (not uncommon) belief that by running (mainly machine learning) algorithms on big data, we can detect patterns and use those to potentially make policy decisions. Here is a cautionary talk:

See the resources section for more examples of this phenomenon.

Your task

Your goal is to produce a 3 minute video (shorter is OK) on YouTube that talks about impact of algorithms (good and/or potentially bad) on individuals, organizations and society. Your video should talk about one specific case study and has to be done in groups of size EXACTLY 3. The best ten videos (as determined by me) will get to present their mini project on last day of class for 4 minutes.

What exactly constitutes a case study?

A case study consists of an algorithm (or a class of related algorithms) that solve a real-life problem. So when you choose a case study you have to pick both the algorithm(s) and the problem that they solve. There is some amount of leeway on what constitutes a class of algorithms but the class has to be fairly specific-- in particular, you should be able to write about the algorithms at the level of algorithm ideas you write in your homeworks. For example these two examples are OK:

Pagerank is a specific algorithm used by Google to solve the specific problem of ranking its search results. This is OK as a case study since both the algorithm and the problem are very specific.
Collaborative filtering is a class of algorithms that is used to recommend movies (e.g. on Netflix ). In this case we have a class of related algorithms that is used to solve a specific problem of recommending movies.

The following are examples that are not OK as case studies (since they are too general):

Machine learning algorithms are used in recommending movies. Here machine learning algorithms is not specific enough.
Collaborative filtering is a class of algorithms that is used in recommender systems. Here the problem of recommending systems is too broad.

What exactly counts as a society?

We will purposefully not define what exactly we mean by "society" to allow for you to make as broad an interpretation as possible. Here are some illustrative examples to set you on the right path. Society could mean part of our system of laws (e.g. the criminal justice system ) or it could mean an occupation (e.g. financial services ) or it could be a certain demographic (e.g. refugees ) or it could mean a part of culture (e.g. Hollywood ).

However, specific organizations and individuals is not OK.

What exactly would be an organization and an individual?

This part should be fairly obvious. When we say organization, we mean named organizations: e.g. Google or Netflix or Buffalo police department .

An individual could be you or some other human being (like Bertrand Might ).

Logistics

We begin with the logistics of deadlines and the various things that have to be submitted. Later we will expand on what each milestone entails.

Deadlines

For your convenience, below are all the relevant deadlines. The deliverables are due by 11:59pm of the due date.

Deliverable	Due Date
Team Composition	September 25
One page pitch	October 4
Video link	November 13

Team Composition and Submission Deadlines

You form groups of size exactly three (3) for the mini-project. Below are the various logistics:

A list of members of your team is due by 11:59pm on Monday, September 25. You must submit for each member their first name, last name and UBIT id. Send this information in an email to atri "at" buffalo "dot" edu.

Note

Your group needs to have exactly THREE (3) members. In particular, I will not assign groups and it is your responsibility to make sure you form a group before the deadline. If you do not know many people in class, feel free to use piazza to look for group members. If you cannot form a group of size three by the deadline, then you get a zero for the entire mini-project.
Every group has to pick a specific case study that shows the impact of an algorithm (or a class of algorithms) on individuals, organization and society. While two groups can pick the same (class of) algorithm that solve (similar) problems, the impacts have to be different for different group. A list of case studies already chosen can be found online. For the purposes of the mini project a case study is a real life occurrence that highlights the impact of algorithms.
A one page pitch (excluding references), which has to be in PDF format, is due by 11:59pm, Wednesday, October 4.
The group will submit a link to a YouTube video that the group has created. The link is due by 11:59pm, Monday, November 13.
Grading

The pitch and video will each be worth $1\%$ and $5\%$ of the grade respectively. All members of the group will receive the same grade.

Note

If a group does not submit a pitch then they will get a zero for the entire mini project.
The top ten videos (as determined by me) will be asked to give a 4 minute presentation in class on Friday, December 8. The presentation is worth up to $1\%$ of bonus points (with the constraint that the total points from the pitch, video and presentation cannot exceed $6\%$ of the grade).
For the mini project there is no restriction on what sources you can use.

Details on your submissions

Next we present what is expected from each of the pitch, video and presentation as well as their grading guidelines.

The pitch

The pitch is a checkpoint for me to provide some feedback on your choices as well as a way to make sure no case study is chosen by more than one group. As mentioned before, the pitch is a PDF document that is at most one page (except the references). Below are some more details on the expected content of the pitch:

Brief description of the case study along with a reference for your case study.

This should include a brief description of the problem, and
A brief description of how the chosen algorithm(s) work: i.e. a brief algorithm idea.

Brief description of the impact of algorithm(s) on individuals in the case study.

Note

To get full credit for this part the individual being impacted should be outside of CSE.
Brief description of the impact of algorithm(s) on organizations in the case study.
Brief description of the impact of algorithm(s) on society in the case study.

Note

Your claimed impact on organizations and society must be backed up by verifiable references.

Hint

If your writeup says something like "algorithm blah should be useful" or "algorithm blah could be useful" then you are probably not doing it right.

Common Mistakes

Many submissions speculate about impact of an algorithm instead of backing the claimed impact by a reference or are not specific about the impact saying something like "algorithm blah has positive impact". Both of these will result in loss of points.
The algorithm idea has nothing beyond a description of the inputs (and maybe the output). You also need to give some idea of how the algorithm converts the inputs to the output. Also if you are trying to "guess" what the algorithm does then you are in trouble.

FAQs

Based on our experience of mini-projects in the last few years, here are some of the FAQs (and our answers):

Can we have a group of size less than three? + -
No. See above.
How do we "reserve" our case study choice? + -

Please submit your pitch on Autolab. The case studies are assigned on a first-submit-to-Autolab-first-get-assigned basis.
Can we have more than one page of text since we have a lot of material to cover? + -

Your pitch will be graded based on the first page of the pitch. Anything beyond the first page (except the references) will be ignored.
Can the impact be based on simulation studies? + -
No. The impact has to be real life.
Does everyone in the group have to submit a pitch? + -
No. See below.
Do the citations have to be in a certain format? + -
No: any format that is understandable is fine. Do put in a URL if possible!

Submission

You need to submit one PDF file to Autolab. Make sure you add relevant URLs (if possible) for references for your case study.

One submission per group

Only one submission per group please. Form your group on Autolab first before you submit your pitch. You can keep re-submitting your pitch till the deadline (with no feedback).

PDF only please

Autolab might not be able to display files in formats other than PDF (e.g. Word cannot be displayed). If Autolab cannot display your file, then you will get a zero (0) on the entire question.

Grading Guidelines

We will follow the usual grading guidelines for non-programming questions. Here is a high level grading rubric specific to the pitch:

Case study description: 20 points. Out of this description of the problem being solved is worth 5 points while the algorithm idea is worth 15 points.
Impact on an individual: 20 points. To get the full points the "impacted individual" must be outside of CSE. You can interview someone at UB who is not in CSE or you can find a reference that specifically talks about impact on an individual. You can talk about the impact on someone in your group but in this case you will not get beyond 10 points.
Impact on organization(s): 30 points. To get full credit your claimed impact must be backed up by a reference.
Impact on society: 30 points. To get full credit your claimed impact must be backed up a reference.

The video

The video should have the same content as the pitch but with more details. Some more remarks:

If you prefer, you can submit an unlisted YouTube video if you would not like your video be public.
By default, your submitted video will be on the CSE 331 webpage. If you prefer that your video not be on the webpage, please note this explicitly when you submit a link to your video.

For the sake of completeness we present the full details on the video part below.

Brief description of the case study along with a reference for your case study.

This should include a brief description of the problem, and
A brief description of how the chosen algorithm(s) work: i.e. a brief algorithm idea.

Brief description of the impact of algorithm(s) on individuals in the case study.

Note

To get full credit for this part the individual being impacted should be outside of CSE.
Brief description of the impact of algorithm(s) on organizations in the case study.
Brief description of the impact of algorithm(s) on society in the case study.

Citations are needed!

Your claimed impact on organizations and society must be backed up by verifiable references. In particular, the citations for your references must appear in the video itself.

Hint

If your video says something like "algorithm blah should be useful" or "algorithm blah could be useful" then you are probably not doing it right.

Common Mistakes

Many submissions speculate about impact of an algorithm instead of backing the claimed impact by a reference or are not specific about the impact saying something like "algorithm blah has positive impact" or the algorithmic impact is not on the chosen problem. All of these will result in loss of points.
The algorithm idea has nothing beyond a description of the inputs (and maybe the output). You also need to give some idea of how the algorithm converts the inputs to the output. Also if you are trying to "guess" what the algorithm does then you are in trouble.
Talking about a case study that was not chosen by the group in the first place. This will result in loss of all points.

FAQs

Based on our experience of mini-projects in the last few years, here are some of the FAQs (and our answers):

Does every group member have to appear in the video? + -
No. In fact, it is fine if your group uses a voice over and no one appears in the video.
Can we use external sources in our video such as pictures or another video? + -

Using external media is fine as long as you put in proper citations. However, all external videos must be short and should be avoided if possible.
Can we use a screen recorder and voice over? + -

Yes. See above.
How do we put in citations in the video? + -
Ideally, you should put in the relevant URL whenever you need to cite it. However, do not expect me to click the link to read up on the details. Your video should be self-contained. In particular, if you are quoting a part of your reference, do it explicitly in the video.
Can we refer to citations on our pitch?+ -
No, your video has to be self-contained. See above.

Submission

You need to submit one PDF file to Autolab. The only thing the PDF needs to have is the link to your video.

One submission per group

Only one submission per group please. Form your group on Autolab first before you submit your pitch. This has to be done separately for the video and the pitch.

PDF only please

Autolab might not be able to display files in formats other than PDF (e.g. Word cannot be displayed). If Autolab cannot display your file, then you will get a zero (0) on the entire question.

Grading Guidelines

We will follow the usual grading guidelines for non-programming questions. Here is a high level grading rubric specific to the video:

Overall video quality: 15 points. The video does not need to be of production quality but it needs to be engaging. Make sure that your video is audible!
Case study description: 15 points. Out of this description of the problem being solved is worth 5 points while the algorithm idea is worth 10 points.
Impact on an individual: 20 points. To get the full points the "impacted individual" must be outside of CSE. You can interview someone at UB who is not in CSE or you can find a reference that specifically talks about impact on an individual. You can talk about the impact on someone in your group but in this case you will not get beyond 10 points.
Impact on organization(s): 25 points. To get full credit your claimed impact must be backed up by a reference.
Impact on society: 25 points. To get full credit your claimed impact must be backed up a reference.

The presentation

Your slides should be emailed to me by 11:59pm on December 7.
Your slides have to be in PPT, PPTX or PDF format.
Your presentation has to have five (5) slides in the following order: (1) Title slide (which must not include any detail other than the title and the group composition), (2) case study overview, (3) algorithmic impact on an individual, (4) algorithmic impact on an organization and (5) algorithmic impact on society.
You will have four (4) minutes for your presentation. The time limit will be strictly enforced. If 4 minutes are done in mid sentence, then you will have to stop in mid-sentence. So make sure you practice to keep the timing under control. It is fine to finish before 4 minutes.
The set of speakers is up to the group. I would recommend having one speaker. This will help with the timing. Of course if you as a group decide to have more that is up to you but remember the above point.
You will be able to use your slides (they will be uploaded on to the computer in the room) and will be able to use a microphone.
In addition to the content, your presentation will also be graded on delivery.

Other Comments

Unlike other aspects of the course, for the mini project you can refer to any source you want as long as (i) you explicitly refer to your source and (ii) the pitch, video and presentation are your own.
Once a case study has been chosen by a group, the choice cannot be changed. Hence, make your choice carefully, keeping in mind that you need to demonstrate the impact of algorithms on your chosen segment of society.
Only one pitch and video should be submitted via Autolab per group.

Resources

Reading material

This reading list on societal impacts of algorithms has a lot of pointers. Happy reading!

Videos

Under Construction

The plan is to put up videos here as we go along. If you find a relevant video, please email me a link!