The Platform

Social media platforms allow users to connect with others. Above all, however, their task is to determine the information that is relevant for us.

• Collect, store and provide information

• Matchmakers: filter content, recommend topics, and suggest friends

Essential questions

• What types of social media platforms do we know?
• What are the incentives from a platform’s perspective?
• How do the platforms influence each other?

In-class task: Reading exercise

Please read the article General Mills, Audi and Pfizer Join Growing List of Companies Pausing Twitter Ads. Afterwards we’ll discuss the following questions in-class:

• What are potential reasons for these decisions?
• What is the role of Elon Musk and especially his other endeavors?

Information flood & the role of similarity

• In cases of massive choice sets decision support is crutial

• We can pre-filter the choice sets based on user preferences

  • Can be predicted based on past behaviour and/or entity charateristics

  • Requires assessment of similarity between entities (e.g., products, items, users)

Short vs. Long-Tail

Two opposing theories:

• Winner-takes-it-all (by Anita Elberse)
  • Digitalization leads to a even higher concentration
• Long-tail (by Chris Anderson)
  • Niche products gain success
  • Superstars loose market share

Content based (filtering)

Content-based filtering uses similarities in products, services, or content features.

• Based on product

• Does not require purchage information (no coldstart problem)

• Often quite similar recommendations

Example: Recommender system

1. Open the spreadsheet (see QR-code)
2. Wait for the instructions (e.g., type of survey, options)
3. Fill it out (one line per person)
4. Calculate user-wise similartiy (by row)
5. Calculate product-wise similarity (by column)

Recap: The vector space

Let $\overrightarrow{x}$ be a vector of size (i.e., dimensionality) $d$:

• Lenght: $||\overrightarrow{x}|| = \sqrt{\sum_{k=1}^{d} x_{k}^{2}}$
• Unit vector: $\hat{x}=\frac{\overrightarrow{x}}{||\overrightarrow{x}||}$

Let $\overrightarrow{a}$ and $\overrightarrow{b}$ be unit vectors of size $d$:

• cosine similarity:

$$cos(\overrightarrow{a}, \overrightarrow{b})=\overrightarrow{a} \cdot \overrightarrow{b} = \sum_{k=1}^{d} a_k\times b_k$$

The cosine similarity takes always a value between -1 and 1; 0 means independence (i.e., orthogonality) of the vectors.

In-class task: Experience online personalization

Take a mobile device (laptops or smartphone), go to the following websites/services and compare what you see with your colleagues.

• google.com; search for “LaTex”
• Open your spotify app (if you are customer) and go to “Your Daily Mix”

Challange: Political bias on social media

Social media platforms like Twitter (more-or-less) carefully position themselves to users, customers, advertisers, and policymakers by making strategic statements.

• Platforms are curators of public discourse
• Challenging to serve all constituencies
  • independence
  • non-partisanship
  • freedom of speech
  • …

Challange: False News

• Political, financial, and social factors are named as the main motives
• False information spreads farther, faster, deeper, and more broadly then true news
• Humans are more likely to be responsible for the spread of fake news than bots

Further reading

• Online Book - Platform
• Pariser, E. (2011). The filter bubble: How the new personalized web is changing what we read and how we think. Penguin.
• Gillespie, T. (2018). Custodians of the Internet: Platforms, content moderation, and the hidden decisions that shape social media. Yale University Press.
• Netflix released a film called The Social Dilemma. It critically examines content filtering and the power of social media platforms that comes with it.