Markets as a solution to the alignment problem
HMIA 2025
2m
SYNOPSIS
Hayek (1974 Nobel Prize): Hobbes is so wrong. Wise Leviathan not possible. Just set ground rules and let agents interact. Order will emerge.
Schelling (2005 Nobel Prize): Market humility is in order. Individual preferences can aggregate to produce collective outcomes contrary to individual preferences.
What were your impressions?
5m
MARKETS EXERCISE
HMIA 2025
AGENT GOAL: trade for a chair that better meets your needs
Chairs vary: color, style, cushionynesss, controls
Each dimension varies from 1 to 5
Each person has a set of preferences [col, sty, cus, con]
My utility for this chair is
\begin{align*}
U(\text{this}) = 16 - (& |\; \text{color}_{\text{pref}} - \text{color}_{\text{this}} \;| \quad+ \\
& |\; \text{style}_{\text{pref}} - \text{style}_{\text{this}} \;| \quad+ \\
& |\; \text{cushiness}_{\text{pref}} - \text{cushiness}_{\text{this}} \;| \quad+ \\
& |\; \text{controls}_{\text{pref}} - \text{controls}_{\text{this}} \;|)
\end{align*}
\text{If my preferences are } (3,3,2,3) \text{ then my }
U(1,1,1,1) = 16 - (|3-1|+|3-1|+|2-1|+|3-1|)=16-7=9
compute utility of your initial chair endowment
repeat
mingle
trade if mutually advantageous
until stopFor example:
10m
5m
2m
HMIA 2025
PRE-CLASS
5m
SOCIETY COMPLEXITY COULD NEVER HAVE BEEN DESIGNED
CANNOT BE IMPROVED
WITH DIRECTION ONLY GUIDING RULES
3m
Takeaways : Hayek
Order can be designed or emergent.
Even emergent order depends on background rules.
Organizations (planned orders) exist within emerging orders.
Government's role is to keep the game running, not to play to win.
Rules, not rulers, gives rise to cooperation.
Order means predictability.
Prices and norms carry local knowledge.
Planning the whole is computationally impossible.
Trying to plan the whole undermines the benefits of local problem solving.
3m
to exist
if otherAgentHere
engageto engage if mutually advantageous exchange
Rules
- Honesty
- Hold up your deals
"What the general argument against 'interference' thus amounts to is that, although we can endeavour to improve a spontaneous order by revising the general rules on which it rests, and can supplement its results by the efforts of various organizations, we cannot improve the results by specific commands that deprive its members of the possibility of using their knowledge for their purposes.
Does the argument against central planning hold up in a world of (practically) infinite bandwidth and processing power? Are there situations where local knowledge is inferior? Where global system knowledge trumps?
DEBATE: THE ARGUMENT AGAINST INTERFERENCE
CLAIM (51.3)
"We shall see that it is impossible, not only to replace the spontaneous order by organization and at the same time to utilize as much of the dispersed knowledge of all its members as possible, but also to improve or correct this order by interfering in it by direct commands."
5m
STOP+THINK
HMIA 2025
PRE-CLASS
5m
Takeaways : Schelling
Not very biased individuals can produce massively biased results.
Stable but undesirable outcomes.
Systems of interacting agents exhibit
non-linear behavior: tipping points, cascades.
Like Hayek: Order can emerge without central control.
Collective results can be undesired by almost everyone.
Individually rational behavior leads to collectively irrational outcomes.
Individual incentives can be misaligned with social welfare.
Global order can be explained by simple local rules and interactions.
Fair, desired outcomes may require institutional intervention, not just better individual values.
Individuals just don't want to be in a minority
unstable equilibrium
stable equilibrium
stable equilibrium
Number Expected
Number Who Come
10m
Social scientists have catalogued a lot of these problems that can emerge from market relations.
MARKET FAILURES
1m
"Market Failures"
Information Asymmetry. Markets fail when one side has more or better information than others. Classic example: Market for lemons.
Externalities .Markets don’t internalize environmental, social, or systemic harms.(Spillover Costs or Benefits)
Market Power / Capture. When some actors can dominate, manipulate, or distort market processes.
Coordination & Public Goods Problems. Markets underprovide goods that are non-excludable or non-rival (e.g., safety standards, infrastructure, basic research).
Systemic Risk / Fragility. Markets prone to contagion or collapse .
Principal–Agent / Incentive Misalignment. Agents (managers, contractors, AI systems) pursue their own interests rather than the principal’s.
3m
What (else) Can Go Wrong?
Information Asymmetry. Markets fail when one side has more or better information than others (classic Akerlof “lemons” problem)
Externalities (Spillover Costs or Benefits).Markets don’t internalize environmental, social, or systemic harms.
Market Power / Capture. When some actors can dominate, manipulate, or distort market processes.
Coordination & Public Goods Problems. Markets underprovide goods that are non-excludable or non-rival (e.g., safety standards, infrastructure, basic research).
Systemic Risk / Fragility. Markets prone to contagion or collapse require safety and robustness measures.
Principal–Agent / Incentive Misalignment. Agents (managers, contractors, AI systems) pursue their own interests rather than the principal’s.
5m
NEXT TIME IN HMIA: Groups
The Emperor's Dilemma
(NAKED KINGS)
Silly
Rules
(BERRIES)
Order Without Law
(COWS)
2m
HMIA 2025 Markets
By Dan Ryan
