Ethical Implications#

Week 11 — What the Index Does Not Price#

Where we left off#

• Ten weeks building a tool for finding the optimal ratio
• Week 10 made it constrained, and therefore usable
• the tool is now good enough to be worth arguing about
• this week: the argument

Learning objectives#

By the end of this lecture you should be able to:

• state what the index’s objective function leaves out
• explain why externalities scale with adherence rather than spend
• hold “the model is correct” and “the outcome is bad” apart
• write a position rather than a disclaimer

What the objective function says#

Week 10 maximises total standard drinks subject to a budget.

Read that sentence literally. It contains:

• a quantity to maximise
• a constraint

It does not contain a person, a timeframe, a consumption rate, or a consequence. Those are not oversights in the formulation — they are simply not in it.

Why that variable, specifically#

• spending more does not concentrate standard drinks per purchase
• following the index does, by construction — that is what it optimises for
• so the externality tracks how strictly the rule is applied
• which means the model’s own success metric is the thing that scales it

This is an uncomfortable result and the reading does not soften it.

Two judgements, not one#

A model can be:

• mathematically correct — the arithmetic checks, the constraint binds, the optimum is the optimum
• describing an outcome that is a poor idea to execute

These are evaluated by different methods and neither settles the other. Ten weeks of this course established the first. None of it establishes the second.

The two ways this argument fails#

• the hedge — “apply the optimisation responsibly”. Names no variable, forbids no outcome, would read identically under any inventory whatsoever.
• the joke — treating the externality as the amusing part of a serious document, which additionally tells the reader not to trust the rest of it.

Your portfolio statement has to survive both tests. The week 11 lab applies them.

What a position looks like#

It names something specific your model does, and says what follows.

At event scale the multiplication is explicit: the externality is multiplied by the guest count, which is a quantity you chose and can defend or revise. That is a claim a shopper-scale statement cannot make.

Common errors#

• writing a disclaimer and calling it a position
• concluding the optimisation should not be taught, which is not what the reading argues
• treating “the index is correct” as settling whether to follow it
• deferring the whole question to the reader

Where this connects#

• the week 11 lab drafts and tests your statement
• the Event Portfolio is marked partly on whether it engages honestly here rather than hedging
• week 12 scales all of it, this included, to fifty people

Next week#

The capstone: what changes when one basket becomes an event.