Quiz3
===

# 3

> In the paper “Using the Lessons of Behavioral Economics to Design More Effective Pay-for-Performance Programs” (Avlok), the article talks about designs to improve the performance of doctors. But they discuss that studies have shown mixed responses to the use of incentives; what are the reasons for say low performance even when periodic incentives are handed out? They suggest using very high incentives, what would be the pitfalls of it? – that is, a doctor is paid huge sums as incentives for every surgery. Explain in terms of behaviour while social context is known.

- Listed reasons as per the paper for the low impact of P4P schemes on doctor performance:
- schemes are constructed haphazardly, in terms of what "seems reasonable" for a healthcare provider to implement
- high thresholds for incentives that put them seemingly out of reach for practitioners
- purported complexity of incentive schemes allagedly make them too hard to account for during each individual decision that a doctor needs to take to win the incentive
- framing effects
- infrequency of payouts
- relative size of a payout compared to salary, allegedly leading to physicians anchoring on the salary amount when interpreting the value of a payout that is paid alongside the salary
- long delays between performance and payout, allegedly leading to time discounting issues
- "teaching to the test", ie. gaming the metric that the incentive is tied to (possibly at the expense of overall quality of care)
- "intrinsic vs extrinsic" motivation conflicts.
- solutions and improvements that paper offers in light of the existing models:
- framing interventions
- break up incentives into smaller pieces and pay out with greater frequency
- reduce payout delays
- decouple payouts from salary
- change payouts to be in kind, not cash, further reducing comparability
- tiering absolute thresholds, so as to improve "participation" of doctors with low baselines on an incentivized metric
- "reducing complexity" by using absolute or quantile thresholds - this suggestion somewhat contradicts the tiering suggestion.
- in general their recommendations seem to be in the line of increasing the frequency of incentive-ralted activity to a doctor's practice, and therefore increase the salience of incentives to a practitioner.
- problems with the offered suggestions:
- it is possible that P4P programmes mismeasure. Given the extreme complexity of the practice of medicine, what is appropriate data to understand *overall* outcomes might not be representative for *any specific* outcome. Therefore attempting to create a virtuous cycle over a general metric by treating it as a score in specific cases is likely to be entirely inappropriate.
- to borrow the statistical idiom of medicine itself, data that is appropriate for analysis at the epidemiological level might be inappropriate for determining diagnosis or care.
- a "danger" listed in the latter section is "extrinsic-intrinsic conflict" - which seems better understood as the converse result of [<a href="#gneezyFinePrice2000">gneezyFinePrice2000</a>] .
- gist of above citation: when fines were introduced for lateness in israeli daycares, lateness rates went *up*. suggestd explanation is that a fine is treated as a price, replacing the extant trust relationship with a transactional one.
- the converse result would be that when *additional* rewards are introduced for someone's performance at some subtask of their jobs, that becomes a price on performing that subtask.
- the other suggested issue is "teaching to the test" - the problem of incomplete contracts as it applies to performance-based pay as discussed in [<a href="#hartTheoryContracts1987">hartTheoryContracts1987</a>] .
- This problem is best illustrated via the concept of an Italian strike, or "work to rule" - a strike wherein workers perfrom their duties *as specified in the terms of their employment* ; this prevents workers being guilty of breach of contract, but results in no work getting done anyway - because the majority of the expected duties of any worker cannot actually be fully specified.
- any metric that attempts to measure skilled work, to which an incentive is tied, is caught between the scylla of intractable complexity and the charybdis of simple inadequacy.
- the above two problems, when understood to apply in the same environment, create a feedback loop wherein an inaccurate measure is being incentivized at direct cost to actual performance. It seems as though the principled approach of a doctor with what this paper calls "extrinsic motivation" - ie. a meaningful commitment to the practice of medicine - would be to ignore this malformed incentive structure altogether.
- a corollary: the principled approach of a mechanism designer or choice architect would be to recommend the scrapping of P4P programmes altogether until a better architecture for them can be found. (this paper instead chooses to discuss how doctor's decision and attention heuristics might instead be hacked in a manner that is agnostic to whether the hacking will cause better care to be provided.)

# 8

> In the presentation by Rohan Gover , there was a discussion that human moral decisions are governed by statistical expectations about what others will do and normative beliefs about what others should do. These vary across different cultures and historical contexts. As per the discussions in the class and your understanding of society (based on membership in it), what varies between cultures to show differences in human moral decisions. Consider the payoffs in a prisoner’s dilemma setting to explain the role of culture.

- Moral behaviour (going by Hume-an construction of "morality") is a collection of drives that are separable from "rationale", ie. drives that don't derive from the drive to be consistent or optimal in achieving one's goals.
- Components of human behaviour that seem closely related to the semantics of "morality":
- prosocial behaviour
- caregiving
- fairness (inequity aversion)
- The above tendencies can be understood to construct "constitutive" norms, as per Searle's usage [<a href="#SearleSpeechActs">SearleSpeechActs</a>], with corresponding "regulative" norms that implement those tendencies in specific contexts, which may between contexts without the underlying incentive structure changing.
- Eg. dissonance aversion - the tendency to prefer one's judgment to align with one's peers - is an individual tendency;
- a related constitutive rule would be the enforcement of a dress code within a specific location;
- one of the corresponding regulative rules is the ban on shorts within the Vatican. It does not significantly differ in incentive structure from other dress codes or behaviour restrictions in other places of worship, even though the specific restrictions may differ.
- An alternate conceptual bifurcation of behaviour around norms: "innate" vs "deliberative", ie. :
- norms, or responses to norms, that arise from lower-level processing
- rooted in evolutionary history
- common across human populations
- norms or responses that arise from deliberative or reflective processes
- involves situational or contextual processing
- likely to be constituted via social interaction, communication, or "politics"
- likely to exhibit clustering within and variance between human populations
- Both these bifurcations seem to cleave along similar lines, though not perfectly overlapping.
- eg. Situational exigencies can lead to constitutive rules that are also deliberative norms.
- notably, prisonner's dilemma games, when contextually embedded, are not invariant to changes in overall resource scarcity, or in the relative size of the stakes.
- if, for eg., the profit achieved in cooperating is *not enough* for the purposes of either player, there is no choice but to attempt a defection. Loosely interpreting "prisonner's dilemma cooperation" as "prosocial behaviour", we would expect to see crime rates correlate with poverty.
- [<a href="#yamagishiTrustEvolutionaryGame2011">yamagishiTrustEvolutionaryGame2011</a>] discusses at length the distinction to be made between enforced cooperation and emergent collaboration, and the the analogous first-person-perspective concepts of assurance and trust. Concretely, the allegation to be made is that in an environment where prisonner's dilemma games are largely played in a oneshot fashion, most players defect; but if they are played in an iterated fashion, more collaboration is possible; and if an *act of trust* can be observed, ie if the PD game is converted to an extensional form where one player is seen to choose to attempt collaboration before the other player needs to choose, then collaboration happens significantly more often.
- [<a href="#graeberDebtFirst50002012">graeberDebtFirst50002012</a>] provides an interesting example of this phenomenon, noting that the widespread usage of coinage often coincides with either war or the slave trade - situations where the likelihood that a given party will be around to settle an account becomes very low, and trust relationships are impossible. Graeber notes the co-occurence of low-trust and the community instability, which may arguably be a causal relationship, or cross-reinforcing one.
- Notably, the latter separation of concerns implies that discussions of the *norms* that human societies adhere to ought not be wholly equated with the idea of *morality*.
- the existence of tendencies that are common to all humans can probably be used to derive a notion of "good" that is general enough to apply to humanity as a whole; a minset of morals, as it were.
- the notion of neuromorality might therefore have important humanistic implications.

# 7

> Can you apply a financial or economic analogy to the trolley question and analyse the outcomes of moral judgements and dual-process theory? (Refer to paper” Beyond Point-and-Shoot Morality:Why Cognitive (Neuro)Science Matters for Ethics “ presented by Murali.

- A hospital currently has one million dollars left in its budget for this quarter. They could spend it on improving the dated ventilation systems, thus reducing the spread of infection on premises and saving hundreds of lives every month; or they could spend it on acquiring equipment for one very complex and expensive procedure for a current inpatient, which will save the life of that patient, and the three-on-average similar cases they see in a year.
- Legal restrictions exist in many countries on the specific ways one can make a decision to trade QALYs against each other in hospital settings. These tend to be sensitive to context and framing.
- The dual-process theory argues that there are two "characteristic" approaches to the trolley problem and its brethren, which correspond to clusters of ethical systems, here called "consequentialist" and "deontological".
- consequentialist\* approaches attempt to find a way to *optimize* on available outcomes, which are arranged into a preference ordering by inducing it from some shared salient details across the outcomes.
- deontological\* approaches attempt to act to best align with a given policy or heuristic, ie. to direct the behaviour itself to align as much as possible with a policy, as opposed to acting to cause a result.
- The paper in question makes a case for these approach-clusters being neurologically grounded - a tension between system-1 and system-2, or between heuristic and logical reasoning; where the balance between the two is what generates most variation in ethical systems or real-world approaches to moral dilemmas.