Price algorithmic harm: the cognitive externality is at least as well-documented as the carbon one

The extraction logic that built American hegemony — unlocking physical resources, the cognitive bandwidth of the workforce, the public-trust commons of broadcast spectrum and telecom — has penetrated the digital sphere. In its current form, it extracts the most valuable cognitive resource a citizen possesses (attention) and dumps the byproducts (polarization, civic-trust erosion, eroded democratic capacity) onto the public commons. By the partial accounting that exists today:

• $78 billion per year in global disinformation losses — stock-market panics, operational disruptions, impaired corporate and consumer decision-making (estimated 2024–2025 annual flow). This is the externality cost the platforms do not pay; the broader economy does.
• $5.6 billion in free earned media for a single presidential campaign (Mediaquant on the 2016 Trump cycle, roughly 3× the nearest rival). The premium proves that the media establishment systematically subsidizes provocation for its own ratings — a structural transfer of value from the public discourse to the engagement-optimization layer.
• A documented decline in democratic capacity.The 2026 Global Social Progress Index reports that the “Rights and Voice” subscore has dropped 6 points globally since 2011 and nearly 2 points since 2021 — a structural decline driven in large part by the algorithmic amplification of outgroup hostility. The United States is one of only eight countries to show a net decline in overall social progress over the past fifteen years.
• Hyper-polarization that prevents legacy legislative work. Political polarization in the United States has been on a steeply rising trajectory since the mid-1970s (NBER). The structural consequence: Congress cannot address the country's biggest legacy problems — the $39 trillion federal debt, decaying infrastructure, the broken healthcare ledger — because the legislative bandwidth required has been captured by the polarizing symbolic controversies that maximize engagement for the platforms that profit from them.

This is the legacy bill of the cognitive externality. The platforms were built to connect people; the engagement- optimization layer accumulated as a workaround that compounded as the business model evolved. The pattern is the same canon framing elsewhere — a century of asymmetric incentives accumulated into the present arrangement. The Accord names the failure, prices it where it appears, and proposes the architectural repair the rest of this paper details.

The engagement-optimization industry is not a cultural accident. It is an infrastructure organized around emotional extraction. Three operational loops compound to produce the measured effect:

• Rage baiting and algorithmic sorting. Platforms do not passively mirror human interest; they actively teach machines to look for emotional heat. When Facebook introduced reaction emojis, its own engineers discovered that “angry” reactions generated five times more engagement than standard “likes” — and the platform proceeded to weight “angry” reactions fivefold in its ranking algorithms. TikTok's internal analytics document that posts with combative or negative tones drive 1.6 times longer watch times than positive ones. These are not bugs. They are commercial optimizations.
• Anger clicking as unpaid digital labor. Every reply, quote-tweet, or stitch designed to correct or mock a provocation feeds the platform's engagement metrics. The act of denouncing outrage measurably amplifies it, because the algorithm interprets engagement as relevance and increases distribution accordingly. Moral concern is converted into monetized engagement at no cost to the platform.
• The physics of perceived importance. A rough but useful formulation captures the inflation dynamic. Symbolic controversies of low intrinsic significance can be amplified into apparent crises by maximizing the volume of outrage expressed about them and the loudness of that expression. Complex systemic issues that don't generate the same emotional heat get crowded out of the perceived agenda even when their cosmic importance is far greater:

The mechanism is not subtle. The platforms know precisely what they are doing; the internal engineering documents surface periodically through whistleblower disclosures and antitrust litigation. The mechanism is also not malevolent in design intent. It is the rational profit-maximizing response to an incentive structure no one has reformed in the twenty-five years digital advertising has existed.

The financial value of the loop is concentrated and the money flows are well-documented:

• Engagement-optimized CPMs. Platforms with the highest engagement-per-user metrics command the highest advertising rates from corporate brands. The economic incentive points unambiguously toward whatever content keeps users active longest. The empirical pattern — Facebook anger 5×, TikTok combative 1.6× — shows exactly what that content is.
• The attention monopoly. Because the cost of expressing outrage online is exceptionally low (minimal physical risk, near-zero marginal cost), the threshold for participation is significantly lower than offline. This produces cheap, continuous, scalable content, which platforms convert into ad inventory.
• The political fundraising loop. Outrage media has become the primary vehicle for political fundraising. High-polarization political actors raise historic volumes from highly targeted lists of anxious donors. The 2016 Trump campaign's $5.6B in free earned media is the single best-documented case, but the mechanism is bipartisan and systemic: provocation maximizes attention; attention maximizes fundraising; fundraising funds further provocation.
• Olsonian distributional coalitions in the digital era. The engagement-optimization industry — platforms, ad-tech intermediaries, the political consultants who depend on the fundraising loop — functions as exactly the kind of distributional coalition Mancur Olson identified in The Rise and Decline of Nations. The coalition coordinates, lobbies for protective regulation (or against constraining regulation), and erects entry barriers against would-be reformers. The aggregate effect is economic rigidity and slower productivity growth — the same pattern the Accord names at /believes/reward-velocity.

Like a chemical processor dumping waste into a public waterway to avoid the private cost of treatment, the engagement-optimization industry dumps cognitive pollution — polarization, civic-trust erosion, degraded democratic capacity — onto the public commons. The cleanup is currently borne entirely by the public:

• Erosion of rights and democratic capacity. The 2026 Global Social Progress Index reports that after a decade of growth, global social progress has stagnated and in many places declined. The “Rights and Voice” subscore has dropped 6 points globally since 2011 and nearly 2 points since 2021. The decline is directly correlated with the hyper-polarization of public opinion driven by algorithmic amplification of outgroup hostility.
• Cognitive sclerosis and divergent factual realities. NBER documents that US political polarization has been rising steeply since the mid-1970s. The polarization is not merely a difference of opinion. It is a divergence in the perception of factual reality — driven by algorithmic curation that systematically biases each user's information diet toward the content most likely to keep them on the platform.
• Legislative gridlock on legacy problems. When the legislature is captured by the engagement- optimized polarization the platforms profit from, the long-term problems that don't generate the same engagement go unaddressed. The $39 trillion federal debt, the decaying infrastructure, the broken healthcare ledger — these are the legacy bills the country cannot address because the political bandwidth has been spent on the issues that maximized advertising inventory last quarter.

The standard regulatory framing for decades treated the internet as a self-regulating, open public square. As the downstream costs of algorithmic manipulation have threatened the basic stability of democratic states, a major global regulatory transition has begun. Two peer interventions are particularly instructive — both shift legal liability from individual users to platform architectures:

Both interventions accept the same architectural premise: digital platforms are not unregulable neutral spaces. They are designed systems that produce documented harms. The state can — and increasingly will — regulate the design.

The primary defense the engagement-optimization industry mounts against intervention is the assertion of constitutional “free speech” protections. The argument conflates two genuinely different things:

• The right to speak vs. the right to algorithmically amplify. Free speech, as a core tenet of democratic capitalism, protects the right of an individual to express a belief or opinion. It does not impose an obligation on society to tolerate, nor does it create a corporate right to construct, highly optimized non-linear amplification engines designed to artificially prioritize incendiary speech for private financial profit.
• Algorithmic prioritization is conduct, not expression. The algorithms that drive the engagement-optimization industry are not neutral conduits of human speech. They are automated sorting machines designed to maximize platform dwell times. The decision to weight an “angry” reaction five times higher than a “like” is a commercial conduct choice aimed at revenue optimization. It is an action taken to exploit a cognitive vulnerability, not an act of political expression.
• The physical-nuisance analogy. A factory owner cannot argue that the right to free assembly or free enterprise permits operating high- decibel machinery that destroys the hearing of nearby residents. A digital platform cannot credibly claim that “free speech” protects the right to operate cognitive feedback loops that destroy the mental health of youth, degrade public discourse, and incite real- world violence.

This is the same content-vs-conduct distinction US courts have applied to physical externalities for a century. Speech is protected. The corporate right to extract value from amplifying that speech, while dumping the resulting cognitive and social damage onto the commons, is not.

The architectural answer is straightforward and rides on existing canon — three pieces already named in the Accord:

• Engine 6 (Externality Limiter), extended. The Accord's canonical list of priced harms — carbon, methane, financial speculation, systemic-risk subsidy, interchange extraction, public-health harms, pavement damage, aquifer depletion, institutional concentration, labor-market undercut — extends naturally to the cognitive and algorithmic domain. Platforms that operate amplification engines producing measurable downstream harm (polarization indices, mental-health outcomes, civic-trust erosion) pay a fee on the harm at source. The mechanism is the same one applied to physical pollution: price the cost where it originates so the market can decide which uses are worth their full bill.
• The Digital Safety Board. One of the Accord's Five Expert Panels (Statistics · Healthcare Quality · Housing Finance · Financial Stability · Digital Safety), the Digital Safety Board is Senate-confirmed, methodology-audited, on staggered terms, and statutorily insulated from administration-level political pressure. It sets the algorithmic-audit standards, publishes the harm-measurement methodology, and triggers automatic stabilizers within statutory corridors. It replaces the discretionary, crisis-by-crisis political bandwidth that produces the gridlock the platforms benefit from today.
• Democracy Hardening (Engine 7) protections. Ranked-choice voting, independent redistricting, real-time campaign disclosure, the 17-day open voting window, the inspector-general protections — the broader hardening agenda exists specifically to protect democratic capacity from concentrated influence. The cognitive externality is one of the largest sources of that influence today. Engine 7 mechanisms make the democratic system structurally more resistant to it.

The Accord's position is not to ban platforms, censor content, or pick which speech is allowed. It is to price the documented harm of the amplification architecture so present-day markets — not future generations — decide which platform behaviors are worth their full bill. The same logic the Accord applies to carbon: charge the cost at source; let the market choose from the options that survive at the honest price.

The engagement-optimization industry represents a classic market failure of exactly the kind the Accord exists to repair: a highly profitable extractive operation that imposes its multi-billion-dollar cleanup costs directly onto the democratic and social fabric of the nation. By treating attention as a costless resource and social cohesion as a free dumping ground, the current platform architecture systematically enriches a narrow operator class while hollowing out the institutional capacity, trust, and rights of the citizenry that uses it.

The ongoing interventions in Europe and Australia demonstrate that the era of treating digital platforms as unregulable neutral spaces is ending. The challenge for the United States is not to suppress dissent or pick which speech is allowed. It is to structurally price the extractive business models that convert human attention and anger into corporate revenue, so the market can sort which forms of platform conduct are worth their full cost.

The closing thesis. A democracy that prices what is dumped on its commons survives the century. A democracy that treats the dumping as protected speech does not. The mechanisms required are not exotic. They are the same externality-pricing and expert-panel architecture the Accord already runs on. The only thing required is to extend them into the cognitive domain, where the harm is at least as well-documented as the carbon damage that Engine 6 already addresses.