AI summary by TwinMind
The trouble with Facebook and what it has in common with scientific publishing
Sabine Hossenfelder
7m 53s
Facebook's Current Problems
- Facebook receives negative media attention for multiple issues
- Collected and passed on user information without consent
- Provided platform for organized spread of political misinformation/fake news
- Creates information monopoly harmful to democracy
- Similar to why freedom of press written into Constitution
- Prevents information monopolies
Historical Context of Algorithm Filtering
- Scientists have written about internet information filtering problems for 2+ decades
- Original concerns focused on search engines as news providers
- Role now taken over by social media platforms
- Core problem: company algorithms select what information users see prominently
- Information selection dramatically affects user opinions
2003 Research on "Googlearchy"
- Key paper by three political scientists coined term "Googlearchy"
- Found web not as egalitarian as expected compared to traditional media
- Empirical results showed enormous disparities in links to political sites
- Small number of heavily linked sites receive more links than rest combined
- Sites effectively dominate their political communities
- Introduced "Googlearchy" - rule of most heavily linked
- Web structure funnels users to few heavily linked sites per political category
- Now evolved to "rule of most heavily liked"
Current Exploitation of Algorithms
- People now deliberately exploit algorithms to misinform others
- Situation much worse than 2003 predictions
- Deliberate misinformation campaigns targeting algorithmic weaknesses
Parallel with Scientific Publishing
- Scientific publishing has similar filtering problems
- Uses citations as primary filtering technique
- Number of citations may be disconnected from real-world research impact
- Citations only demonstrate popularity, not scientific relevance
- Citations as disconnected from scientific relevance as likes are from truth value
Key Differences Between Platforms
- Social media has little tradition of quoting sources
- News outlets copy from each other, making accuracy verification difficult
- Social media has much faster turnover rate than scientific publications
- Less time for users to think before passing on information
- Both suffer from near-monopoly of single algorithm
Proposed Solutions
- Problem stems from users unable to customize filtering algorithms
- For scientific publishing: make alternative evaluation methods more accessible
- References website "a metered org" as solution example
- Scientists currently use citations/journal impact factors because easily accessible
- Other evaluation metrics not easily accessible
- Journal impact factor calculation methods unknown to most users
Facebook Algorithm Transparency Issues
- No one knows how Facebook's algorithm works
- Algorithm cannot be customized by users
- If users could customize information filtering, gaming would be less problematic
- User customization settings should remain private information
User Customization Implementation
- Most users wouldn't want to deal with technical details
- Sufficient if small group creates templates for user selection
- Personal example: uses Facebook for science news and close friends
- Doesn't want political news, unknown people's social lives, or entertainment
- Other users have different needs: groups, events, distraction, cooking recipes
Current Facebook Limitations
- Very little customization options for newsfeed
- Best current option: sort people onto lists (cumbersome, limited solution)
- Only solves some aspects of sorting problem
Proposed Facebook Solution
- Allow third-party plugins to sort newsfeeds
- Would give users more control
- Would relieve Facebook of some responsibilities
- References Zuckerberg's motto: "move fast and break things"
- "Unless you are breaking stuff you are not moving fast enough"
- Time to break Facebook's dominance over information filtering
