Proposing ‘libertarian paternalism’ - designing choice architectures to influence people to make better decisions for themselves.
Libertarian paternalism. Paternalism - want people to make decisions which lead to best outcomes, as judged by them. Libertarian - don’t want to restrict freedom to choose for self. The libertarian part is interesting as an escape hatch - if a particular policy is really bad it can be averted by informing individuals rather than waiting for slow moving government to catch up.
Choice architecture - context in which choices are made.
Nudge - any aspect of choice architecture which would influence a human but not an Econ ie does not significantly change incentives but does change behavior through irrationality or bias.
There is generally no neutral context, no way to not influence choices. If any action is a nudge, you have a responsibility to consider the effects. For example, a common nudge is the choice of default decision if the person in question abstains, forgets or can’t be bothered. Should set a default that benefits the person.
Much focus on nudges in government, because there are many important nudges that the government is in the best position to influence eg healthcare.
Not for bigger government, for better governance. Already bipartisan support for nudges in savings.
People are willing to pay other people to help them make better decisions (eg support groups for quitting smoking) - evidence that they do not have full free will over all of their actions. Nudges interact with the portion that they don’t control, so they don’t reduce freedom to act.
Popular libertarian strategy is to maximize choices. This only works when the chooser is well informed, rational and attentive. Can also cause choice overload, resulting in worse decisions.
Humans and econs
Biases and blunders
Alternate names for System 1 / System 2 - automatic system and reflective system. Not nearly as good as Albert and Rex.
Typical list of biases and experimental evidence. I notice that some of the examples given are effectively using nudges for Evil.
If snacks are in reach you are likely to eat them. You might be grateful if someone helpfully took them away. An Econ would never benefit from having a choice removed.
Dynamically inconsistent - preferences which do not agree over time. Right now I want to wake up early. Tomorrow I will likely want to sleep in. Behavior like this cannot be understood in terms of a single set of preferences.
Another alternate set of names - Planner vs Doer. Albert and Rex! Memorable characters stick in peoples heads. Planner and Doer do not get along. Planner sets you alarm clock for 6am. Doer turns it off and goes back to sleep. Doer is the main victim of nudges. Also Doer is sneaky. Can often convince Planner that it was Planners idea all along.
Reasons why Doer often wins. Temptation - call a choice tempting if we are more likely to make it when in the heat of the moment (hot) than when thinking ahead (cold). Mindlessness - much of the time Planner is asleep at the wheel and the irrational Doer is making the choices. Generally not aware of mindlessness either.
Many strategies for Planner to influence Doer. Pre-commitment - remove choices in advance. Social pressure, incentives - speak Doers language.
Mental accounting - fungible assets often partitioned into multiple accounts eg rent money, food money, vacation money. Doesn’t make sense to an Econ, but acts as an aid to self-control. Understanding this is helpful for many nudges.
Following the herd
Social pressure is powerful eg obesity is contagious. Can divide into conflating popular choice with safe/optimal choice vs worrying about negative opinions if not conforming. Can use for nudging eg “Don’t mess with Texas” campaign significantly reduced littering.
Conformity experiments. Actual perceptions can be overridden by majority opinion. Strongest when opinion is unanimous - even single dissenter can weaken the effect.
Herd behavior, winner-takes-all effects, fashion effects - rerunning popularity contests (eg choose favorite song under peer influence) can produce random winners. A single confident confederate can sway the entire group. Arbitrary preferences are internalized and can persist even years later. Appears in voting where noone wants to waste a vote by voting for someone that will not win - if you can create a perception of one candidate winning you may increase their votes.
Has interesting implications for public debate. Lacking concrete evidence, simply repeatedly and confidently stating your position could have huge effects.
Pluralistic ignorance - often widespread ignorance of other peoples thoughts/opinions. Explains how unpopular regimes can be stable for a long time, and then suddenly crumble if some stimulus causes everyone to realize that everyone else also hates the regime.
Heavily influenced by peers - implies you can make major changes to your own life by carefully choosing your peers. This is a good part of why I’m considering a postgrad course in cogsci - I don’t think the lectures and material add much value over reading alone but being surrounded by other students and staff may well be valuable.
Strategies: information, peer pressure, priming.
Information - people influenced by group norms. If their view of group norms is inaccurate, can influence their behavior with correct information eg heavy drinkers hang out with heavy drinkers so have a biased view of normal drinking levels, giving them correct stats reduces their drinking. Both extremes move towards the norm, so may want to selectively target people if you only want movement in one direction.
Peer pressure - concerned about other peoples opinion. Spotlight effect - believe that people pay more attention to than they really do. Seen elsewhere - choosing popular students to promote new behavior is much more effective than choosing well behaved students.
Priming - subconscious influence by spreading activation eg posters with eyes make people behave better, as if they were being watched. Mere measurement effect - simply asking someone whether they will perform an action makes them more likely to act in accordance with their answer eg asking people if they are going to buy a car increases buying rates. From the description, presumably the people who answered no were then less likely to buy a car?
When do we need a nudge?
Nudges are most important for decisions that people get wrong. What attributes make decisions hard?
Requires self-control - Planner struggles.
Difficulty - requires a large investment of time/effort to make the correct decision.
Infrequent - practice makes perfect, but many important decisions happen rarely in a lifetime eg buying a house.
Slow feedback - long time between decisions and feedback makes people less likely to make the connection.
Unclear feedback - not clear whether or not a good decision was made eg you got fat - which of the thousands of things you ate in the last few years contributed?
Complex - can be hard to map choices to futures.
Lack of self-knowledge - even knowing how choices map to futures, humans are poor at predicting which future will make them happy.
Free markets are not a panacea. Ways they can fail people:
Insufficient competition - monopolies or collusion tend to lead to abuse.
Lemon markets - if people cannot sufficiently judge quality they tend to get ripped off.
Irrationality - there is an incentive to exploit any bias, irrationality or lapse in self-control.
Phishing for Phools takes this further and argues for a corollary of the Efficient-market hypothesis - that where human flaws exist, they will be exploited.
Stimulus-response compatibility - if stimulus/affordance does not lead Rex to correct response, Albert has to intervene, reducing performance and accuracy. Eg big red go sign, push door with pull handles.
NUDGES - iNcentives, Understand mappings, Defaults, Give feedback, Expect errors, Structure complex choices. TODO come up with an acronym that I will actually remember.
Incentives. Who uses? Who chooses? Who pays? Who profits? When incentives are misaligned, people get screwed. Incentives might also be out of sight - salience can be improved by putting information on display eg increasing price of electricity will be more effective if accompanied by regular reports of how much each appliance is costing the owner.
Understand mappings. Map choices to things the user values eg medical choices can be mapped to costs, outcome stats, inconvenience. Use units that are familiar in the context eg ml of alcohol vs pints of beer. RECAP - record, evaluate, compare alternative prices - proposed regulation that would require companies to export pricing and usage info in a form that comparison websites can use to make recommendations. This is by far the most interesting suggestion so far - it doesn’t clash with ideals on either side and it would be politically difficult for companies to object to.
Defaults. Often not changed, so provide powerful nudges. Required choice - avoid default, insist uses chooses something. In absence of knowledge, choice is unhelpful - do I want my monitor to run at 60hz, 70hz or 80hz?
Give feedback. Acknowledge actions. Warn about potential errors - but be careful of false positive exhaustion. Make things easy for Rex to learn eg provide affordances, don’t move buttons around, prefer regular schedules (eg birth control packs include placebo pills for days when no hormones should be taken). Make sure that feedback is actionable eg if input is wrong, explain what needs to be corrected and how.
Expect errors. Try to make incorrect actions impossible, or handle them correctly anyway. Warn about common errors, and notice if they occur. Reduce the impact of errors. Post-completion error - once main goal is accomplished remaining tasks more likely to be forgotten - if necessary delay completion eg ATMs wait till card is taken before handing over cash. I have occasionally taken the card and walked off without the cash.
Structure complex choices. Provide a choice process. Elimination by aspects - pick most important attribute, pick cutoff values, remove all options that fall outside cutoff, repeat until number of choices is manageable. ML recommendations are good example, but beware of filter bubbles.
Save more tomorrow
US personal savings rate at lowest since Great Depression. Most people do not save enough to retire safely.
Should we interfere with their choices? Surveyed employees mostly responded that they ‘should be saving more’. In many cases they are literally giving up free money. They spend little time on the decision of how much to save. Typical savings amounts are often round numbers or equal to the maximum employer matching rate, indicating that little thought/planning went into the decision. Interventions that increase savings rate see no increase in dropout rate. So, on the whole, probably people could do with a nudge.
Automatic enrollment. Required choice (at hiring time). Required choice with a recommended option. Enrollment online. All increase participation. Also suggests raising matching rate (while keeping maximum matched amount constant) to a sensibly high round number.
Expanding number of options decreases participation. Financial education does not increase scores on financial literacy tests and has only a small effect on enrollment.
Save more tomorrow - allow employees to precommit to raising their savings rate on every pay increase. Sidesteps loss aversion, takes advantage of future virtue being easier that virtue now. Substantially increases savings.
Pension Protection Act. Incentivizes employers to provide automatic enrollment, increase contribution rates over time and match employee contributions.
Deciding when to cash out social security is tricky and online calculator is useless.
Pension funds have to be allocated to different investments. It’s complicated, there is a risk-EV tradeoff and the allocation needs periodic adjustment.
Loss aversion means that risk tolerance depends on how often one looks at the results.
Most people are heavily risk averse (which is likely a mistake) and tend towards buying during bubbles and selling during crashes.
Many people, including some named economists, simply split the money evenly between all the offered funds. Means that the choice of funds directly influences the risk-EV tradeoff.
Many people invest heavily in their employer, believe that it is no risker than a diversified fund and base their valuation on previous performance of the stock. Regulation around the running of pension funds don’t apply to company stock (thanks to lobbying) so there is an incentive to push them on employees. Suggests ‘Sell More Tomorrow’ - give employees option to automatically sell off company stock over time - avoids crashing the price or sending negative signals while still divesting employees of unnecessary risk.
Defaults. Now that the DoL has issued official guidelines on how much risk is advisable, companies can safely default to that level rather than the most conservative option, which tends not to produce enough return to actually retire. Can also offer multiple defaults eg conservative, moderate, aggressive.
Structuring complex choices. Provide a tiered choice - a single default -> a few reasonable plans -> full control.
Expect error. Automatic enrollment, defaulting to Save More Tomorrow. Offer an automatic rebalancing to those who took manual control.
Mappings and feedback. Map plans to expected lifestyle, plus estimate of risk.
Incentives. Enforce laws that require pensions to be managed in best interest of employees. Remove exemption on company stock.
Mortgages used to be simple. Legislation required standard format for reporting interest rates. Much more complicated choice now, opening the door to both better and worse choices. Study finds that higher costs are predicted by race and education, and that brokered mortgages are more expensive. Typical result - complexity results in less sophisticated buyers being taken advantage of by advisors.
Subprime mortgages - high interest rates for risk loans. Provides chances to people who need them, but also opens avenue for exploitation through confusing and manipulative sells. Complicated paperwork often hides fees and legally-required warnings.
Lauren Willis proposes limiting types of mortgages to reduce complexity and adding compulsory delays between receiving estimates of total cost and signing. Author proposes RECAP instead, plus simplified report that reduces costs and fees to single number. Side benefit is that it makes shopping online easier, which reduces discrimination against minorities. (Also notes that RECAP strategies could succumb to collusion between comparison sites and sellers.)
Student loans. Government (need-based) or private. Need-based forms are complicated and difficult. Private loans are easier and have less restrictions - appealing to unsophisticated teens. Some college loan officers are in collusion with private companies. Combination of loan guarantee and government subsidy makes student loans very profitable, but competition has focused on rent-seeking instead of prices. Phishing for phools! Government created a market with unsophisticated buyers and zero risk of default. Blood in the water.
RECAP. Simplify government forms and require same questions for private loans. Make standard report format for all loans. Combine government form with tax return to save on duplicated effort.
Nudge families into college savings plans.
Credit card debt is growing. Estimate average family spends $2k on interest per year. Credit cards break mental accounting - bottomless jars. Rewards encourage Rex to spend. Minimum payment anchors actual payment - encourages higher interest fees. Often no way to default automatic payments to pay the whole bill.
RECAP report should prominently include total cost of using the card (interest + hidden fees), in addition to making it easier to shop around for a better deal.
Privatizing social security: smorgasbord style
Swedish government privatized Social Security. Went full pro-choice - 783 funds. Adverts (both government and private) encouraged ditching the default and picking your own - 2/3 did.
Default portfolio is well-designed. Average citizens chosen portfolio (does averaging over the portfolios make any sense?) is higher-risk (mostly stocks), under-diversified (mostly Swedish), has higher fees and requires more active management. Default fund has consistently outperformed the average choice. Very few people have modified their portfolio since creation.
Fund adverts contain little actual financial information, implying rent-seeking and exploits.
Ideally author would have had them offer tiered choices, along with advice that the default option was carefully constructed by experts.
Prescription drugs: part D for daunting
Part D - add choices to medicare. Well intentioned. Offered structured choice and filtered options. Mostly opt-in, but low-income seniors and disabled were required to enroll, and defaulted to a random plan if no action was taken. Growing penalty for late enrollment. 50+ plans. Massive public encouragement to pick a plan. Donut hole - coverage turns off after a certain amount of spending and turns back on after another threshold.
Medicare and hospitals swamped with support needs. Much confusion. Some advice bodies also added their own plans - conflict of interest. Many people still unenrolled, including a large subset for which subsidies add up to make it practically free.
Random defaults = sometimes making choices we know are suboptimal - there is no neutral default. Also removes competition - participation nets an equal share.
Poor UX on website. Takes hours to fill out questions. Four experimenters, given same data to fill in, all ended up with different answers.
Recommends removing the law that forces random assignment.
RECAP. Just before enrollment opens, send patients a standardized list of their medication use over the last year.
Estimates that correct plan choice would save an average of $500/year. Study showed that sending a personalized price comparison is a somewhat effective nudge to pick the cheapest plan.
How to increase organ donations
Choice between opt-in, opt-out, or not option at all (some states allow taking corneas without consent). Switching from opt-in to opt-out increases donation rate 16% (according to what sounds like a complicated analysis controlling for many factors).
Author prefers opt-out but with easiest possible interface. Another alternative is mandated choice eg to renew drivers license, must check yes/no for organ donation. Advertising with mild social/moral pressure is also effective.
Saving the planet
Environmental concerns typically addresses by strict regulation. Avoiding tragedy of commons usually requires some form of Leviathan.
Author prefers pricing externalities directly. How are carbon credits working out? Approach is generally not popular because it makes the high costs of fixing the problem very explicit. Suggests using mental accounting eg using taxes on carbon to fund universal health-care. Such a program for reducing acid rain was a huge success. How might I judge this for myself? The author has a clear bias.
Feedback. Mandatory labeling has worked for many areas (eg pesticides, cigarretes). Disclosure requirements for industrial toxins have also been very effective in reducing toxic releases - environmental groups used the reports to target PR campaigns against the worst offenders. Again, little detail. Is it worth following up the reference to judge for myself? New labels for cars show annual fuel cost as well as mileage - suggests adding social pressure by putting the labels on the bumper for all to see.
Ambient Orb - little ball that glows red when owner is using lots of energy. Lead to 40% decrease. Could add social pressure by making the results public.
EPA has engaged companies in voluntary programs to reduce energy costs - setting out standards and allowing any company that meets them to use the corresponding logo. Provides leverage - it’s much easier to pick the machine with the sticker than to research the energy usage of each.
Improving school choices
School choice is tricky. Allow private schools and rich kids get advantages. Disallow them and parents lose freedom over their children are educated. Income-adjusted subsidies might undercut public schools.
Study finds that public schools forced to compete produce higher student achievement.
Attempt in Worcester to do something RECAP like is foiled by schools sowing doubt and obstacles in the path of transfers.
Most parents send their children to the default school. Choice booklets don’t contain much standardized information. Condensing the stats to an easily comparable form on a single sheet led to low-income families making similar choices to high-income families, where before they were less likely to apply to highly performing schools.
Complex selection systems, designed to aid de-segragation, are gamed more often by higher-income, more educated, more connected families. In Boston a group of economists designed a replacement system that incentivizes honesty.
Simple nudges like helping students fill out forms and bringing admissions tests to schools can have large effects.
Should patients be forced to buy lottery tickets?
Litigation is a significant cost to hospitals and that cost must be passed down to customers somehow. Estimated to account for 5-9% of hospital expenditures. Also indirect costs in the way it alters behavior to avoid litigation risk. You can also sue the NHS, but it only accounts for 1% of their costs and those costs are already half that of the US. The NHS directly insures doctors rather than requiring a 3rd party insurer. Are the two cost estimates not comparable? Do US patients sue more often? Do the malpractice insurers add to the cost (via profits, interaction time, cash flow)?
Argues that patients might want to give up their right to sue, in exchange for not bearing the costs.
Doctors pay more or less the same insurance premium regardless of how many times they are sued, so suing does not incentivize their behavior. Does discipline board not penalize doctors at all? Why is insurance sold like that anyway? There is very little correlation between whether review determines that negligence exists and whether the patient sues and whether the claim pays out. Huh? Who is running this system?
If we regard the right to sue as insurance for the patient, it looks like very expensive insurance. Suggests that health insurance plans should come with the choice of whether to be allowed to sue, to make the costs explicit. Also proposes that patients should only be allowed to sue only for reckless or intentional wrongdoing, not for negligence. Why is the doctor being sued anyway? Generally sue companies, not employees. Are doctors not normal employees?
Some countries have adopted a no-fault system where you get compensation for medical injury regardless of the cause, saving the high costs of lawyers and insurance companies. Injured parties get compensation. Hospitals are incentivized to lower injury rates. Might lead to not wanting to take on cases where injury is likely eg by refusing to specialize in such fields and directing patients elsewhere.
If we disentangle the religious and cultural aspects from the governments involvement, what we are left with is an array of legal and financial rules that apply only to pairs of people who enter into a government contract together. Different tax rules, entitlement to certain employment benefits, shared ownership of property and funds, surrogate decision making and additional privileges in court. Not to mention visas - a guarantee of being able to live together regardless of future employment situation.
Aside from the cultural significance, not substantially different to any other government system eg tax avoidance programs, attorney-client relation. Like a government incentive to commit to stable child-rearing, which then accreted other tangentially unrelated laws. Not even a particularly effective way to protect children vs eg mandatory child support, high-quality free childcare. Question is not ‘should we protect children’ but whether this is the most effective way to do it or whether we need to tie it into so many other systems.
I think a much stronger argument is from the baseball-bat-to-the-head principle - if marriage didn’t already exist, would we invent it?
…oh, this gets mentioned one the next page :)
People who are about to marry generally don’t believe there is any chance of getting divorced, so don’t opt-in to pre-nups, leaving them with the messy default divorce law. Would be much less messy if the default set of outcomes was much clearer eg fixed formula for child support rather than court decision.
A footnote ducks the issue of polygamy, which would have made for a much more interesting chapter (eg http://qntm.org/gay - is marriage transitive?). Dealing with those kinds of issues would be much easier if marriage were carved apart at the seams as the author suggest, allowing each group to consider exactly what commitments they want to make to each other.
Extensions and objections
A dozen nudges
Give More Tomorrow - like Save More Tomorrow but for charitable donations. This is exactly how I reached the popular 10% mark - never triggered loss aversion.
Charity Debit Card - tracks charitable donations for easy tax returns.
Automatic Tax Return - anyone without itemized deductions or unreported income receives a pre-filled tax return which they can sign and return. In the UK you don’t even have to fill in a tax return if you don’t want to make any claims. There is also a tiered architecture, where the official online form expands sections in response to questions. A similar system just launched for small companies. It now only takes me around an hour per year to do both personal and company taxes.
Stikk.com - public pre-commitment with financial or social penalties for failure.
CARES - smokers deposit every week what they would have spent on cigarettes. After six months, if a urine test confirms they have not smoked, they get the money back. More effective than nicotine patches.
Motorcycle helmets - instead of mandating, require a stricter driving test to be allowed to ride without helmet.
Gambling self-bans - online gambling sites in the UK are legally obliged to implement and enforce self-bans and self-limits.
Destiny Health Plan - health insurance with reward scheme for activities that reduce health risks.
Dollar a day - teenage girls with a baby receive a dollar for every day that they are not pregnant.
Change lights for aircon filters
Nobite - nail polish that tastes bad, preventing nail biting.
Disulfiram - drug that cause alcholics to throw up immediately if they drink - breaks reward cycle.
Civility Check - “Warning: this appears to be an uncivil email. Are you sure you don’t want to wait a while and think about it?”
Slippery slope. If only implement each individual policy if its widely beneficial, is there a slippery slope at all? Libertarian - low-cost opt-out. Some nudge is inevitable anyway, might as well choose a good one.
Evil nudgers. Public policy especially might be hijacked by partisan interests. Author views this as an incentive problem. RECAP and similar could be reflectively applied to policy design.
Right to be wrong. Doesn’t prevent people being wrong, just makes it easier for them to be right.
Anti-redistribution. Some believe that all charity should be left to private entities. I don’t really understand this view, and the author doesn’t do much in the way of devils advocate, so basically just beating on a straw-man. Would like to hear arguments from someone who actually holds this position - suspect they would be less nonsensical from the horses mouth.
Imposes costs on Econs. Might actually remove costs from Econs eg reducing load on healthcare and welfare by helping people make better decisions.
Also, sometimes people prefer not to choose. Give them the freedom to give up their freedom!
Publicity principle - don’t allow policies that could not be defended in public, even if they benefit the public. Quick test of whether something is a nudge or manipulation is whether or not you would be happy to explain it to the subject. Also needs to be monitorable eg subliminal messaging might benefit people but is hard to notice if it is crossing the line if you can’t see it.
Neutrality. Tricky to define. Certainly couldn’t nudge people towards some religion, even if it empirically increases health and happiness. Nudging makes less sense in domains where judging what a fully-informed version of the individual would want is harder eg religion, relationships. So limit it to cases where it is clear what people want (at least in abstract) and where there is no conflict of interest (eg no nudges on who to vote for). Transparency and reporting can reduce the number of cases where conflicts of interest can hide.
Generalises to asymmetric paternalism - weighing the benefit of helping the least sophisticated against the costs on the most sophisticated. Libertarian is a special case where costs are near zero.
Mandatory cooloff periods (eg for divorce).
The real third way
- Nudges are everywhere and can drastically influence behavior
- It’s possible to preserve freedom of choice while still influencing behavior for the better
A potential Third Way - something that both Left and Right can agree on.
Twenty more nudges
‘I double’ - establish group norm of avoiding long, drawn out arguments and replacing them with the challenge ‘I double’, at which point the other party can either commit to a standard bet or back down. Works only when there is an unambiguous answer. Still, I’m interested in ideas like this for reducing pervasive and contagious overconfidence.
The entire paper seems to be an extend motte-and-bailey around what exactly is meant by preserving freedom to choose. If a subconscious nudge causes 50% of nudgees to choose the desired option, did they still have free choice? What if it was 70%? 90%?
Freedom itself is such a slippery concept that its barely possible to use as a means of weighing decisions. Many restrictions of freedom open up other freedoms - very few libertarians would object to restricting the freedom to rape and pillage. Many freedoms enable deleterious downstream effects eg if people are free to sell their kidneys, they are free to be economically coerced into doing so. Similarly, if people are free to not purchase health insurance then there is less political pressure to ensure that everyone can afford it.
However, that motte-and-bailey might be worth quietly ignoring if it turns out to be swallowable by both sides of the US political spectrum. Bipartisan cooperation would enable fixing a lot of pressing problems.
Let’s instead ignore the problematic concept of freedom of choice and instead view the book through this lens: how can we influence peoples decisions for their own benefit, without their objection and without the system of influence being subverted by other interests. If people don’t believe that nudges can influence them so strongly then they won’t object to their use. Commercial advertising is far less restrained, both morally and technically, so I don’t see an argument that would allow adverts but frown on nudges.
There are a couple of different kinds of problems covered: Planner vs Doer conflict, lack of information, lack of expertise, bias / lack of self-knowledge. It’s only really on the last problem that the moral questions really come into play - in all the others it’s quite clear what people would want if fully informed.
For Planner vs Doer conflict we have pre-commitment, cooloff periods and speaking the Doers language. For lack of information we have RECAP and feedback, and for lack of expertise we have guidance on systems of defaults and transparency against collusion. I would expect all of these to be largely acceptable on both sides of the spectrum eg the majority of gun owners support waiting periods. RECAP and transparency measures are also hard for companies to publicly oppose (what have you got to hide?), although they would undoubtedly lobby the shit out of them behind the scenes.
For bias / lack of self-knowledge, things are much murkier. If someone doesn’t want to wear a motorcycle helmet is that a reasoned assessment of the risks or the result of inability to recognize their own fallibility. Do we have the right to overrule their decision in either case? I don’t see a knockout argument in either direction.
I don’t have much to take away from this - the actual concrete techniques are all things I’ve seen elsewhere, mostly in HCI texts - except for the idea that you can take advantage of the fact that people don’t believe in their own irrationality. Advertisers and politicians have abused this for decades, so we might as well try to use it for good too.