Networked Markets, MS&E 233



Spring 2012-13, Stanford University
Time: Tuesday, Thursday 2:15pm - 3:30pm
Location: Classroom 370-370

STAFF

Instructors: Ashish Goel, (Stanford, Email: ashishg AT stanford DOT edu, Phone: 650 eight-hundred-fourteen 1478),
Mukund Sundararajan (Google, Email: mukunds AT google DOT com)

Office hours for instructors: Tuesday 4pm - 5pm. Location: Huang B016.

Course Assistants: Hadi Zarkoob (Email: hzarkoob AT stanford DOT edu ), Riley Matthews (rileym1 AT stanford DOT edu)

Office hours for CAs: Hadi Zarkoob, Friday 4pm - 5pm, starting from April 12, 2013. Location: Huang B016.

Staff email address: Please send emails to staff mailing list, msande233-spr1213-staff AT lists DOT stanford DOT edu, unless you specifically need to contact one of us.

Piazza: Please register in the course page at Piazza using your Stanford email address (click on the "Enroll in Course" botton at top right of the page). Once you logged into the course page, you can go to the Q&A section to post questions. Use Piazza to post questions that could be interesting to other students in the class, and please monitor it frequently so you can see other questions that are being asked and answered.

Auditors: Please sign up to the list msande233-spr1213-guests at the Stanford List Server , and request access to piazza by writing to the TAs.


SYLLABUS (SUBJECT TO CHANGE)

We will discuss, via a series of well motivated examples, some of the main principles underlying the design and operation of networked markets. A detailed list of topics is given below. There are no textbooks. We will make notes, slides, or handouts available within two days of every lecture. The course load will be 3 HWs (a total of 30 points), a midterm (30 points), and a project (40 points) to be done in groups of 3-4. In addition, students can attend and submit a brief report (100 words or less) on any two RAIN seminars (http://rain.stanford.edu) during the quarter for up to 5 extra credits. If you receive more than 100%, we will give you an A+.

Topic

Lectures

Details

Introduction

1

Illustrative Examples: Groupon + eBay + craigslist + Ashley Madison + oDesk + rhapsody + netflix + Mechanical Turk. Technical example: Heavy Tails

Auctions

4

The Vickery auction and eBay; Extensions to Sponsored Search (i.e. search advertising); Pricing models for selling ads on the Internet; Reserve Prices in auctions; Designing markets for economic efficiency and budgets; Additional examples of businesses built on search or search advertising

Reputation Systems and Collaborative Filtering

3

PageRank; Reputation Systems based on Incentives and Penalties (eg. eBay's reputation system); Collaborative Filtering; Applications to advertising and retail

Tracking on the Internet, Personalization and
Privacy

2-3

Anecdotes; database definitions; differential privacy; industry standards; cookies; browser tracking; benefits of user specific signals.

Midterm

1

70 minutes; In-class

Long tails and networked models

2

Deeper understanding of heavy tails, including connections to networked models; virality and social marketing; network models

Advanced topics in reputations and markets

2

Credit networks and virtual currencies; bitCoin; peer-to-peer file sharing; prediction markets

Advanced topics in networked pricing

2

Experimental frameworks; pricing; ad exchanges

Project presentations

2

Attendance mandatory



TIMELINE

The midterm will be on May 7, in-class. The project presentations will be on the last two days, May 30th and June 4th. Attendance on the presentation days is mandatory; if you can not make it to those days, please contact the course staff in advance to receive a remedial assignment.


PROJECTS

Please form a group and submit your project preferences by April 11th. Some project ideas are given below. We will add more, if needed. Also, feel free to suggest a different project. We will soon start a mechanism to help students find each other. When you submit your project preferences, we will ask you to submit a ranked list of two projects, and also a ranked list of the main interests of each group (eg. mathematics, economics, business insights, algorithms, software development, data analysis, case studies, advertising, retail).

A midterm report for each project will be due on May 13th. This should be at most four pages, and briefly outline your broad approach, your current progress, future plans, and any questions you have for the instructors. We will schedule 15 minute meetings with each group to discuss the midterm report. The final reports will be due on May 28th, and should be at most 15 pages in length. In addition, each group will be asked to present a 10 minute overview of their project.

  1. Run a campaign for five weeks for a charity or local business of your choice using Google and one out of Facebook or Twitter. We will provide advertising coupons. We will balance teams between FB and Twitter, and at most eight teams are allowed to pick this project. You will be expected to provide insights into how social and search advertising differ.

  2. Design (but do not implement) an incentive scheme for users of the EteRNA project, and discuss its pros and cons. Use pre-existing data to identify good designers and good predictors, and use that to see if you can improve the selection process for the experiments that are actually conducted. Professor Rhiju Das from the BioChemistry department is available for a brief consultation session and for providing access to data. Two scenarios of interest are:
    (A) suppose there are 10,000 participants, and there are only 8 designs synthesized per week. This was the case for eterna in 2011-2012, and will likely hold for any new citizen science projects that follow EteRNA's template.
    (B) suppose there are 10,000 participants, and there are 2,000 designs synthesized per week (this is the current case).

  3. BitTorrent is a Peer-to-peer network for data exchange. BitCoin is a new cryptographic currency. Design (but do not implement) a protocol that uses BitCoin as a virtual currency to make sure that users of BitTorrent are appropriately and fairly incentivized to upload file-blocks. Address as many systems and incentives issues as you can think of, so the output of your project is something that a developer can code. Adam d'Angelo, the CEO of Quora, who conceived of the project, will be available for a brief consulting session. At most two teams are allowed to pick this project.

  4. Help design and run a user trial (in class) for consensus building using widescope. Developers of widescope (http://widescope.stanford.edu) will be available for consulting. At most two teams are allowed to pick this project.


HANDOUTS

Lecture 1 Slides: High-level Overview
Lecture 1 Notes: Long Tails
Lecture 2: eBay, Auctions
Lecture 3: Incentive Compatibility, Revenue of the Vickrey auction, Sponsored Search
Lecture 4: Sponsored Search Auctions: k-item auction, prominence, pay-per-click
Lecture 5: The Incorrectly Generalized Generalized Second Price Auction, Reserve Prices
Lectures 6-7: Reputation Systems and PageRank
Lecture 8: Applications of PageRank to Recommendation Systems
Lecture 8 Supplementary material: Movies in class
Lectures 9-10: Network Models
Lecture 11: Privacy (Part I)
Lecture 12: Privacy (Part II)
Lecture 13: Prediction Markets
Lecture 14: Choosing Between Alternatives


HOMEWORKS

Homework 1
Homework 2