Ezra Klein recently wrote in The New York Times, republished by the AFR, that algorithms are reshaping our attention in ways that matter for our autonomy, wellbeing, and collective life. His central argument is simple: platforms reward content that is hard to look away from — not content that is good for us.
This resonates with my own analytics.
When I publish emotionally charged content, my reach skyrockets. When I share balanced, carefully reasoned thought leadership, visibility shrinks dramatically.
For example:
- Post: AI set to become the most powerful discriminator of our time Impressions: 33,124
- Post: _Framing DEI differently using Maslow’s Hierarchy of Needs _Impressions: 2,553
Same platform. Same audience. Same writer. The difference is emotional intensity.
As it turns out, this isn’t anecdotal.
The Neuroscience Pathway: Attention Prioritises Emotion
Decades of neuroscience — especially Joseph LeDoux’s work — show that the human brain processes emotionally salient information through two distinct pathways:
- The Low Road Fast, automatic, amygdala-driven. Built for survival. Triggered instantly by novelty, threat, moral outrage, or anything emotionally charged.
- The High Road Slow, reflective, cortical. Handles reasoning, nuance, analysis — the kind of thinking thought leadership requires.
📎 https://www.simonandschuster.com/books/The-Emotional-Brain/Joseph-LeDoux/9780684836591
Online environments constantly stimulate the low road:
- sensational headlines
- moralised content
- conflict
- identity threat
- emotionally charged political debate
This creates reliable spikes in arousal and attention.
The AI Pathway: Emotion -> Attention -> Amplification
Social media algorithms learn to reward whatever humans reliably attend to.
This is the critical bridge:
Our biology shapes our attention; algorithms optimise for our biologically driven behaviour; and the result is an ecosystem where high-arousal content dominates.
A growing body of research support this.
1. High-arousal content spreads faster
A landmark MIT study analysing 126,000 Twitter news stories found that false stories — typically more surprising and emotionally charged — spread farther, faster, deeper, and more broadly than true ones.
Key findings:
- False stories are 70% more likely to be retweeted than true ones.
- Truth takes six times longer to reach 1,500 people.
- Falsehoods reach a cascade depth of 10 20 times faster than facts.
- False stories diffuse across more unique users at every level.
Researchers attribute this to human attraction to novelty: false stories tend to be more unexpected, triggering surprise and disgust, while true stories evoke sadness, anticipation, or trust.
Although novelty alone does not cause retweets, it gives falsehoods a structural advantage: emotionally charged information travels first and travels furthest.
📎 MIT summary: https://news.mit.edu/2018/study-twitter-false-news-travels-faster-true-stories-0308
📎 Full paper in Science: https://www.science.org/doi/10.1126/science.aap9559
2. Emotionally expressive content is systematically amplified
A study in Nature Human Behaviour shows that emotional signals — especially anger and moral outrage — are amplified through a three-part process:
- Expression amplification: Emotional posts get more engagement.
- Perception amplification: Viewers overestimate how intense those emotions are.
- Collective amplification: People assume the whole group feels that same level of emotion.
This gives rise to:
- a belief that hostility is normal,
- emotional contagion, and
- escalating cycles of outrage.
Separate research shows that when users receive positive feedback for outrage, they express more outrage in future posts — a reinforcement-learning pattern shaped by algorithmic incentives.
The result: platforms become more emotional than the people on them.
📎 https://www.nature.com/articles/s41562-023-01582-0
3. Engagement-based ranking boosts divisive content
A 2023 audit of Twitter’s recommender systems found that the algorithmic feed systematically amplified political content that was:
- emotionally intense
- hostile toward out-groups
- rated by users as making them feel worse about others
And critically: users did not prefer the content the algorithm showed them.
When researchers tested a preference-based feed, political hostility dropped — but ideological echo chambers deepened.
This reveals a central tension in platform design:
- Engagement-based systems maximise emotion and division.
- Preference-based systems improve wellbeing but risk narrowing worldviews.
Both approaches force society into tradeoffs we have not fully reckoned with.
📎 https://arxiv.org/abs/2305.16941
4. Anger spreads faster than joy
A study of 70 million Weibo posts found that:
- Anger is more contagious than joy
- Anger travels farther through weak ties, reaching new communities
- Anger spreads faster and peaks earlier than other emotions
Negative emotion is simply more “viral.” And algorithms amplify whatever is viral.
📎 https://arxiv.org/abs/1608.03656
Implications for Human Beings
Klein makes a profound point:
Attention is the mechanism through which the brain decides what matters. When algorithms steer attention, they participate in shaping identity, judgement, emotion, and worldview.
Platforms exploit the gap between our instincts (fast, emotional) and our aspirations (slow, reflective).
The consequences are deep:
- Autonomy: Our attention is pre-curated—we see what is triggered by our primal emotive pathways rather than our conscious preferences for higher-level content that supports reflection, growth and learning.
- Wellbeing: Emotional overload affects stress and cognition.
- Knowledge quality: Nuance becomes nearly invisible.
- Public discourse: Outrage crowds out complexity.
- Social cohesion: Bridge-building content has low reach.
- DEI risk: Polarising identity-based content spreads quickly, while inclusive perspectives struggle for visibility—reinforcing harmful stereotypes and deepening divides.
- Pressure on creators: Engagement incentives reward emotion over insight.
- Impact on young people: Their identities form in high-arousal digital spaces.
Implications for Leadership and AI Ethics
If platforms amplify the brain’s low-road responses, then ethical AI must:
- suppress automatic amplification of high-arousal content
- prioritise user wellbeing over engagement
- offer transparent, user-level control of what gets ranked
- elevate educational or reflective content
- assess developmental and societal impact
- reduce reinforcement loops that reward outrage
This is not censorship. It is alignment with human neurobiology and human values.
A Final Reflection
My own LinkedIn analytics show how strongly algorithms reward emotional intensity over thoughtful insight. But the issue is bigger than reach. It touches our collective capacity for:
- critical thinking
- balanced understanding
- respectful dialogue
- long-term decision-making
Ezra Klein is right: we need a shift in both algorithmic design and public expectation.
Attention is not a trivial metric. It is a fundamental human resource — and it determines who we become, individually and together.
Related Reading
Deepfake Abuse Is a Growing Form of Gendered Harm — and Australian Workplaces Need to Respond