Concept
Engagement
Engagement names a metric that becomes a moral language. It becomes important whenever institutions need a neutral-seeming account of what audiences or users want.
Engagement names a metric that becomes a moral language. It becomes important whenever institutions need a neutral-seeming account of what audiences or users want. It helps preserve feedback and product learning, but can fail when response volume replaces judgments about importance, quality, or civic consequence. In When Incentives Become the Moral Language, engagement marks the domain where connection aims survive while platforms learn to say users engage with this content instead of we think this content is valuable—behavioral data looks neutral because it describes user actions rather than editorial preference. It differs from attention because attention names the scarce capacity itself; engagement names the scored response that capacity produces.
Recognition signals
- platforms say users engage with this content instead of we think this content is valuable
- feeds surface what holds attention; what spreads is treated as what the platform believes worth surfacing
- internal research on divisive amplification is set aside when the engagement cost looked too large
- content moderators review material engagement logic sent wide while told they are not editors
- users notice attention calibrated to keep them present rather than informed
Questions to ask
- When does behavioral response stop being a signal and start being a definition of worth?
- What original aims—connection, access, expression—does engagement fail to measure?
- Who carries knowledge the organization cannot say without destabilizing its neutrality defense?
Counterbalances
- retain traceable human explanations for ranking decisions where scale allows
- weigh trust and understanding signals beside raw engagement in product decisions
- name structural deflection when the system promoted what people engaged with
Trajectory
Early signals
- engagement signals are genuinely informative for product improvement
- editorial judgment becomes a legal and reputational liability at global scale
- behavioral data looks neutral because it describes user actions rather than platform preferences
Intensification
- engagement becomes an alternative moral language replacing editorial claims
- anger, fear, and tribal solidarity generate strong response without earning trust
- proposals to reduce emotionally amplifying content are weighed against user-growth targets
Failure modes
- moral responsibility for what the attention economy builds has no institutional address
- the gap distributes across billions of interactions with no single renunciation moment
- mission statements still name connection while operational language is engagement
Restoration paths
- admit editorial culpability where neutrality architecture prevents unilateral fixes
- internal teams elevate amplification concerns even when engagement cost is unpredictable
- users gain language for the structural mismatch between stated aims and feed behavior
Manifestations
leadership
- executives testify that algorithms reflect user relevance signals, not editorial preference
- former product managers describe ranking choices that shaped political behavior under oath
organizations
- ranking engineers raise amplification concerns weighed against user-growth targets
- moderation teams draw lines while distribution decisions stay engagement-driven
software
- recommendation systems optimize behavioral response without measuring improved understanding
- what holds attention gets more attention regardless of mission statements about connection
family
- a household notices feeds surface content that does not match what they said they wanted
