AI summary by TwinMind
Dear Facebook, this is how you're breaking democracy | Yael Eisenstat
TED
13m 30s
Background and Career Transition
- Former CIA officer/diplomat shifted focus ~5 years ago from global security threats to domestic polarization
- Lost ability to engage with non-like-minded Americans on hot button issues
- Discussing divisive topics with fellow Americans became more difficult than engaging suspected extremists overseas
- Left speaker feeling embittered and frustrated
- Feared extreme polarization becoming greater threat to democracy than foreign adversaries
Foreign Service Experience and Counter-Extremism Work
- Served as foreign service officer in Kenya after 9/11 attacks
- Led "hearts and minds" campaigns along Somalia border
- Built trust with communities susceptible to extremist messaging
- Spent hours drinking tea with anti-Western clerics
- Dialogued with suspected terrorists
- Engagements began with mutual suspicion but never resulted in shouting/insults
- Sometimes worked together on areas of mutual interest
- Most powerful tools: listen, learn, build empathy
- Found people wanted to feel heard, validated, respected
Current Social Media Manipulation Problem
- Online environment much harder problem to tackle than overseas extremism
- Current information ecosystem manipulating users
- Entrenching people in absolutism where compromise becomes dirty word
- Social media companies like Facebook profit from segmenting users
- Feed personalized content validating/exploiting biases
- Bottom line depends on provoking strong emotions for engagement
- Incentivizes inflammatory and polarizing voices
- Finding common ground no longer feels possible
- Platforms won't change enough on their own despite growing chorus demanding change
Government Regulation Necessity
- Governments must define responsibility for real-world harms from business models
- Must impose real costs on damaging effects to public health, public square, democracy
- Won't happen in time for U.S. presidential election
- Even with future strong rules, will take everyone to fix
Facebook Employment Experience (2018)
- Hired to head elections integrity operations for political advertising
- Had no illusions about fixing everything but wanted to try steering in better direction
- Didn't work directly on polarization but examined most divisive/exploitable issues
- Russia's 2016 tactic was exploiting divisive issues for election interference
- Asked questions to understand underlying systemic issues
- Title and job description changed on second day
- Cut out of decision-making meetings
- Biggest efforts to combat disinformation and voter suppression in political ads were rejected
- Lasted just under 6 months
Facebook's Structural Problems
- Thousands of passionate employees believe they're making world better
- Company only tinkers with content policy/moderation margins
- Doesn't consider how entire machine is designed and monetized
- Will never truly address platform's contribution to hatred, division, radicalization
- Would require accepting built product might not be best for society
- Would need to alter entire product and profit model
Business Model Issues
- No way to reward listening or encourage civil debate in engagement-optimized business
- User growth and engagement optimization are two most important success metrics
- No incentive to help people slow down or build friction
- No mechanism for people to stop, recognize emotional reactions, question assumptions
- Lies more engaging online than truth
- Salaciousness beats fact-based reasoning in frictionless virality world
- Algorithms feed poison playing to worst instincts and human weaknesses
- Anger, mistrust, fear, hatred amplified (not new in America but dramatically tipped scales)
Facebook's Internal Knowledge
- Recent Wall Street Journal article exposed internal 2018 Facebook presentation
- Specifically pointed to company's algorithms growing extremist groups on platform
- Algorithms polarizing users
- Company knows but engagement drives revenue
Algorithmic Manipulation Tactics
- Modern information environment profiles and segments users into narrow categories
- Perfects personalization process
- Bombards users with information confirming views
- Reinforces biases and creates belonging feeling
- Same tactics terrorist recruiters used on vulnerable youth (smaller/localized pre-social media)
- Ultimate goal: persuading behavior
Proposed Solutions Facebook Could Implement
- Stop amplifying/recommending conspiracy theorists, hate groups, disinformation purveyors
- Stop using same personalization techniques for political rhetoric as for selling sneakers
- Retrain algorithms to focus on metrics other than engagement
- Build guardrails to stop content going viral before review
- Can do without becoming "arbiters of truth"
- Won't go far enough without being forced
- Markets keep rewarding them, not breaking current laws
Current Legal/Regulatory Gap
- No U.S. laws compelling Facebook/social media companies to protect public square, democracy, elections
- Decision-making on rules and enforcement ceded to for-profit internet company CEOs
- Creates post-truth world where toxicity/tribalism trump bridge-building/consensus-seeking
Call for Government Action
- Governments must step up and protect citizenry
- No single magical legislation fix but governments can/must find balance
- Protect free speech while holding platforms accountable for societal effects
- Insist on transparency around recommendation engines
- Transparency needed for curation, amplification, targeting processes
- Hold companies accountable not for individual posts but for how recommendation engines spread content
- Accountability for how algorithms steer people toward extreme content
- Accountability for how tools target people with harmful content
Public Action Items
- Pressure government representatives to stop ceding public square to for-profit interests
- Educate friends/family about online manipulation
- Engage with non-like-minded people
- Make this issue a priority
- Need whole-of-society approach
Message to Facebook Leadership
- People using tools exactly as designed to sow hatred, division, distrust
- Company not just allowing but enabling it
- Positive stories don't make harmful effects okay
- Getting worse heading into election
- Biggest potential crisis if election results not trusted and violence breaks out
- 2021 "we know we have to do better" response no longer acceptable
- Civil rights leaders, academics, journalists, advertisers, employees all warning about harmful policies/practices
- Company owns decisions, can no longer claim couldn't see it coming
