<header class="header">
<h1>Choicemaking and the Interface</h1>
<h3><b>Joe Edelman</b> 2014</h3>
</header>

## Abstract

<p id="abstract">
*Theories of choice from economics and philosophy suggest information requirements for good choices. In view of these requirements, we can see why current menus lead toward regrettable and isolating choices. We propose principles for redesigning menu systems, and consider the possibility of a public database about choices and outcomes. Such a system would reshape the media ecosystem and incentives for business.*
</div>

## Contents

[Introduction](#situation)

Human choicemaking

* **[Searching for Options](#addiction)**
  *Search costs and desperate choices.*
* **[Information Requirements](#life-in-review)**
   *The role of the device.*
* **[Identity](#expressive-choice)**
  *Our values, our choices, our selves.*

[Repairing individual choicemaking](#repair)

* **[Visible Promises](#promises)**
  *What do we hope for when we click?*
* **[Collective Wisdom](#collective-wisdom)**
  *Public data, satisfying choices, & lives well lived.*
* **[Promising Recommendations](#recommenders)**
  *Repairing feeds and autosuggests.*
* **[Value Browsing](#neighborhoods)**
  *Administering our own identities.*

[Beyond individual choicemaking](#beyond)

* **[Social Choices](#social-choice)**
  *Choices are rarely for ourselves.*
* **[Best-Outcome Economies](#outcome-economy)**
  *Consequences for business and society.*

## Introduction

Much of our daily lives are structured by menus. Whether we are deciding which email to respond to, what to order at a restaurant, which stores to visit on a street or in a mall, what job listings to investigate--we navigate our lives by scanning lists of options.

We can ask, then, what are the responsibilities of menu-makers?

This essay focus on one key responsibility: to present options on-screen in a manner that is free from bias and manipulation.  Currently, many kinds of menus (including, on devices--app stores, notification screens, browsers) are guilty of two broad kinds of manipulation:

**Hidden costs**.  *Time costs* for instance are often unexpected on the internet.  Whereas in commerce we are wary of hidden costs, unexpected fees, and consider them *dishonest business*, unexpected costs via the internet are not yet seen as dishonest.

<div class="sidenote">
What if time costs were estimated for everything?
</div>

**False promises**.  When an app store sells us a fitness app or a dating app, are user reviews enough information to keep us from wasting our time? We'll show that, without providing additional information about how they work out for people, app stores mislead us about the promise of a product.  We'd call a salesman a *charlatan* or a *quack*, but we don't yet blame the app store.

This paper takes the position that devices and their menus should serve users. As humans, we tend to choose from what's in front of us. It follows that these menus which are so often in front of us have a great influence on our lives.  We must bring them into line.   At the platform level, devices must take care to populate menus with options aligned with the users' interests--rather than in the interests of increasing engagement metrics for media companies--and should present those options in a way that lets us choose well, lets us express what's important to us, and doesn't isolate us in the process.

In other words, we need a menu environment in which to make informed, expressive, and social choices.  To begin to design such an environment, we'll need an understanding of (a) how our environment affects our choices, (b) how our choices and our identities evolve together, and (c) what aspects of choice our devices could help us with.

## Human choicemaking

### Searching for Options

*"The sea is endless when you are in a rowboat."* -- Casares

Our information environment is the medium through which we search for and compare various options when we are making a choice.  Because of the environment, some options may be much easier to discover or to recall than others.  When an option is hard to find or hard to recall at the moment of choice, we say that it has a high *search cost*.

Sometimes it is our real preferences that determine our choices, but often it is a difference in search costs.  When an exhausted worker comes home and immediately switches on the TV, or a tourist goes to McDonald's because they don't know any local places, or a hungry person buys a donut as a snack when they would, in the abstract, prefer mozzarella and tomato hors d'oeuvres--these can be considered rational choices. But they are made based on the disparity of search costs, not because of preferences independent of search costs.  We can call them *desperate choices*, because they were made in a situation where further search felt too costly compared to choosing something suboptimal.  For a society, they can be an indication of a failing. ✱  <div class="sidenote">✱ To the extent that coordination is an information and search problem, even the choice to do isolated activities instead of meeting with friends is of this type.  For instance, the choice to, on any given day, use internet porn over having a relationship likely is more about search costs (i.e., the difficulty of finding a partner for that day) than about independent preferences (prefering porn to real sex) or non-search transaction costs (e.g., costs of commitment/switching, being charming and humane, etc). </div>

The economics literature suggests that even the most hardcore addictions can be considered as desperate choices.  While popular psychology regards addiction as a lapse in the ability to choose (e.g., "OMG I just spent 4 hours on facebook."), economists tend to think of it as either (a) a rational choice made in a bad situation (e.g., "Life sucks let's get high and watch cat videos.") or as (b) a rational choice made with a misunderstanding (e.g., "I thought I could spend just a few minutes!" or "I thought it'd make me popular and happy!").

These interpretations were advanced in [a landmark paper by Becker and Murphy](http://www.jstor.org/discover/10.2307/1830469?uid=3737864&uid=2&uid=4&sid=21104100113327) which found that cigarette smoking and heroin use are sensitive both to announcements of future tax levels and to independent improvements in life circumstances.  That is, people do stop doing heroin when they suddenly have an unexpected chance to go to college, and they can stop smoking if they figure out it's going to be more expensive than they thought. Similar results have been observed with rats and morphine [1](http://www.sciencedirect.com/science/article/pii/S0014488607000052) [2](http://www.utexas.edu/news/2013/01/23/socially-isolated-rats-are-more-vulnerable-to-addiction-report-researchers) [3](http://www.huffingtonpost.com/johann-hari/the-real-cause-of-addicti_b_6506936.html).

This suggests that a regrettable choice is a desperate choice: not necessarily a character defect, a failure to choose, or a weird preference, but a lack of options at hand.  The activities we regret are the ones where we know we could have done better by ourselves (at least if we had known), but the search for better options was too hard at the time.  ✱  <div class="sidenote">✱ It follows that interventions that deprive an addict of their drug--in the case of internet addicts, browsing "speedbumps"; for alcohol, prohibitions; etc--will not be as effective as providing better situations for people, or placing better opportunities in the menu at the moment of choice.  Behavioral economists are starting to call this [navigability](http://papers.ssrn.com/sol3/papers.cfm?abstract_id=2577018).</div>

Television, the Internet, and now mobile phones have dramatically lowered the search costs for certain kinds of entertainment--mostly isolating entertainment--while giving us confusing menus that don't contain the best options we could think of if we tried.  When we are tired, absent-minded, time poor, or alone, we are likely to choose from the menus given, rather than engaging in a costly manual search.  We are likely to make a desperate choice.

### Information Requirements

*"Of all the words of mice and men, the saddest are, 'It might have been.'"* -- Kurt Vonnegut

But what information in particular is important to guide search, and how can we set an attainable standard for what devices should show us?  How could a device be said to have made a best effort to align with our true interest?

I propose the following: when there is information that would help us *live better* according to *our own judgement and values*, we would like our devices to surface it promptly and at the moment of choice.  I.e., if a device omits easily-obtainable information and so leads us to do something we would regret on review, we may say it has failed us and kept us from choosing well.

To be clear, avoiding the *feeling* of regret is not the standard here.  We often feel regret about an action even when we were fully informed going into it; we sometimes don't feel regret simply because we haven't had the opportunity to review our activity; finally, we may feel regret over a missed chance that we wouldn't have actually preferred.

For this reason, we'll focus on designing interfaces that avoid *durably regrettable choices*--choices where, if we were to review our activity, we'd be able to name a better option that we wish we'd taken (and that we have since become familiar with)--and specifically the subset that we might call *"I wish I had known X" regrettable choices*--where we can also name some plausible information that would have helped us understand the options and choose differently. ✱<span class="sidenote">✱ A full explanation of why "Durable I-Wish-I-Had-Known Regret" (DIR) is a good foundation for defining good and bad choices is beyond our scope here.  But consider that, in choosing a metric around which to characterise choices, we would want one that: (a) accommodates many value systems across different users, (b) aligns with user intuitions about what's good, and (c) aligns with human needs.  An inferior metric like "maximal moment-to-moment happiness" fails all of these:  some people don't value happiness, many things widely regarded to be good like grad school and childbirth are not about happiness, and there are ways to be happy that don't align well with human needs.  Furthermore: (d) DIR is clearly and quickly evaluable via user reports, (e) that evaluation once made is durable over time, and (f) minimizing DIR correlates well with existing metrics, including maximizing brand trust and net promoter, maximizing positive user sentiment, and minimizing burnout.</span>

Shitty menus are those which lead us directly to these regrettable choices.  We may feel Durable I-Wish-I-Had-Known Regret because of wrong assumptions about the costs/likelihoods of the option we choose, about the existence or proximity of better options, or about what is important to us about the choice.  In particular, shitty menus may be missing information about time costs, money costs, whether hoped-for outcomes are likely, whether unexpected outcomes are likely, whether there are less-costly or otherwise better options for similar outcomes, and even whether our hopes themselves are likely to change.

Without this information, it is more likely we'll make a choice we'll later regret.  Below I'll present designs for browsers, for app stores, and for other device menus that address this.

### Identity

*"There are goals other than well-being, and values other than goals."* - Amartya
Sen, 1984

One of the trickiest sources of regret is when we made a choice based on ill-informed values or goals.  For instance, we have all made choices based on price where later we wished we had considered another factor as more important.  Or sometimes we make purchases based on the mistaken idea that we will continue with a certain project.

This happens all the time. Perhaps a goal was overly specific, was dependent on some other goal in a way we didn't understand, was founded on a misunderstanding, or was just less compelling than a goal or value we hadn't thought of yet.

<div class="sidenote">
In this map, an arrow is drawn when people tend to migrate from one activity to another activity, and the latter activity is reported to satisfy their desires more effectively.
</div>

In which situations do our values shift in this way? And how can we make sure that users remain in charge of their own values, rather than having them directed by the device?  As philosopher Ruth Chang has argued, it seems likely that we define and adjust our values while making choices. Specifically, we evolve our values when we have an initial difficulty making a choice because one option seems better in some ways, and another better in a different way.  Dr. Chang suggests✱ <span class="sidenote"> ✱  It would be good to have better data about these moments of value reconsideration.</span> these are the moments where we reexamine and change our values, and that we often choose *a new, more general value* for ourselves that makes the choice more clear.

If this is correct, we need a name for the environments in which this activity of self-construction can proceed fluidly.  We will call these *environments of expressive choice*.  To make an expressive choice, users must be able to explore a space of values that might apply to the choice, and to organize the options *in front of them* in terms of whatever they decide *matters to them*.  As simple a thing as providing widgets for changing the sort order and filtering of options creates more possibility for expressive choice.

If menus are to help us find decisions we won't regret, they must allow us to explore what's important to us about the options, and to navigate according to what we've decided is important.

## Repairing individual choicemaking

### Visible Promises

*"Ads sell a great deal more than products. They sell values, images, and concepts of success and worth."* -- Brene Brown

Without an idea of what the user hopes for with a choice, it seems impossible to surface information about better options or related values.  We will start our redesign with the notion that when the user has decided to engage with a resource--to go to an event, to use an app, to buy a product, to follow a link, etc--it's because she has a hope, because she is taken with the *promise* of the engagement.  This promise could be the *outcome* she hopes will result, it could be a *value* embodied in the choice itself, or it could be an outcome she wishes to avoid.  ✱  <div class="sidenote">✱  Our everyday hopes, as we buy things, download apps, book appointments, etc--are few enough that they can all be clearly named. While human pursuits in general are of infinite variety, the key categories for everyday life are: (a) terms related to ideal schedules (e.g. "frequent exercise"), (b) ideal personal states (e.g. "getting hitched"), and (c) ideal tactics (e.g., "acting boldly").</div>

We may, for instance, think of facebook for a quick break, of Sports Authority for not getting fat, or of voice lessons for a courageous move.  We can give each such hope a name and then use an on-screen object to let the user navigate by what they're really after.

#### Example: Browsing

Most of us have visited facebook.com at some point during work when we needed a break.

Let's say we had been trying to get organized in Gmail.  We don't want to think very hard about what kind of break to take.  We lack the time or the energy to make an internal catalogue of options for a break, so we type 'f' into the box at the top of the window.

In today's browsers, there is no representation of what I want or of how much time I want to spend.  I'm hoping to relax, to take a break, perhaps imagining it will take 5 minutes.  But nowhere on the screen do the words "break" or "relax" or "5 minutes" appear.  It's as though the device wants me to forget my plans.

A more user-aligned browser would make this explicit.  For instance, with an extra tag in the URL bar to capture what the use of the site is about.

Once I start navigating to facebook, of course there could be many reasons I'd want to go there.

Here, I can compare how well these uses of facebook have worked out for people.  And I can see--maybe for the first time--if there might be a better activity for feeling relaxed.

#### Example: Shopping

Of course the same can work for physical goods:

<!--
#### Prior art

The closest thing to a standard for conveying benefits are the [highlighted phrases](http://en.wikipedia.org/wiki/Collocation) pulled out of reviews by Yelp, Amazon, and Google.

The phrases capture *features* of a product or service, like "great view" and "mozarella sticks", but they don't capture our actual hopes, which are both more general and more specific-things like "quality time with friends", "an adventurous weekend", "feeling refreshed", "getting fit", or "learning to knit".

-->

### Collective wisdom

In order to display these kinds of cues, we'd need to build giant database about people, about their choices, and about the outcomes of those choices.

This is a kind of metadata about URLs.  When you think about it, it's similar to other kinds of metadata about URLs that are already in use today.  For instance, Facebook and Google know how many people clicked on a URL or linked to a URL, but they don't currently know how many are glad they went, and why.

We can think of this "big data" about people and URLs as a kind of collective intelligence we share with one another, but when you look at the metrics that we mainly use to propel media coverage--likes / clicks / views / purchases / downloads--a better name might be "collective impulse", because this data is dominated by samples from  the impulsive, lost moments in people's days.

As such, this new big data about people, choices, and outcomes would constitute a real collective wisdom: an alternative to virality, a new claim as to what's worth sharing.

If collective wisdom can be used to drive distribution and ranking of options, it will mean less-regrettable options at hand for most people.  That will mean fewer desperate choices and more lives well lived.

Here's how we might collect it:

An outcomes-based review must remind its user of the costs involved in the engagement (time and money, ideally as authenticated via the service provider or the platform), then give them a quick, precise way to report what the engagement was *about*, how their desired outcomes have worked out so far, and what else has happened.

Even a small number (say [50,000](/OpenOutcomes/)) of reviews could be a significant public resource, guiding people away from likely regret and towards time well spent. To provide meaningful signals across all the top apps and web products, we will need an articulate, representative, and thoughtful set of reviewers, but not so many of them.

As a data API, this collection would be a google maps or OpenStreetMap for life decisions--augmenting all manner of shopping, browsing, and list experiences, allowing users to sort by likely outcomes, relevancy to their own desires for personal growth or for scheduling, etc.

Such an API could be used for sourcing and sorting suggestions for friend get-togethers, for instance:

<div class="sidenote">
This interface was tested with a few thousand users as part of a defunct startup. The results suggested that user behavior can change significantly when recommendations are ordered by desired outcome and ranked by reported likelihood of an outcome working out.
</div>

As a destination site, this collection would be like Wikipedia [or Quora](http://www.quora.com/Life-Lessons/What-is-the-one-thing-that-people-wish-they-had-known-about-life-when-you-were-in-your-20s-d-2), but focused on life choices made, what they were about for the individuals involved, and how it all worked out.  Visitors could review choices about downloading/using apps, websites, products, etc--but they could also review larger engagements, like moving to a new city, going freelance, etc.

### Promising Recommendations

Algorithmic menus like Facebook Newsfeed and Google Now are even more important to align with our true interests, because they've begun to structure even more of our discretionary time.

*A full treatment of recommender systems and feeds will come later.*

<a name="neighborhoods"></a>
### Value Browsing

*To be written.*

<!--
<img src="img/fb-w-lozenge-and-popup.jpg" class="zoomed">

How do people change their activities as they learn about what's possible or what satisfies them?  When the same user reviews different engagements for a set of desired outcomes, we can answer important questions. In the map below, an arrow is drawn when reviewers tend to migrate from one activity to another activity, and the latter activity is reported to satisfy their desires more effectively, so they stick with it.

Many such maps are possible, creating an opportunity for data visualizers and storytellers [to map our collective wisdom](/OpenOutcomes/). What do people that want to "be successful" or "get fit" tend to want 6 months or a year later?  Which desires are the most durable--are you making a big choice based on fleeting interests?  What works best for getting fit--apps, gyms, trainers, fitness buddies, or equipment?  What have people tried first and then given up on?

-->

<a name="beyond"></a>
## Beyond individual choicemaking

*"Isolation underpins technology, and technology isolates in its turn; all goods proposed by the spectacular system, from cars to televisions, also serve as weapons for that system as it strives to reinforce the isolation of 'the lonely crowd.'"* -- Guy Debord

*"Computers, video, radio, printing presses, synthesizers, fax machines, tape recorders, photocopiers -- these things make great toys, but terrible addictions. Finally we realize that we cannot "reach out and touch someone" who is not present in the flesh."* -- Hakim Bey

### Social Choices

*To be written*

<!--

It appears as though we might decide “I’m going to this event” or “I like this video.”

The playground reveals that our choices are usually provisional, invitational, or suggestive, rather than truly personal. Our choices are actually of the form “I will if you will” or “why don’t we” or “here is an example” or “I’ll join you in” or “this I believe in.”

To choose an activity on the playground is to lead a friend, or to hope for a friend, or to invite a stranger with your mere presence, or perhaps even just to enter into a reverie with some equipment or the landscape. The act of rapport and the act of choice cannot be other than simultaneous and it is a conceit unique to consumerism and our tech landscape that we make choices without realizing each choice is as much a planning of others’ lives as of our own.

On some days we may play with the authors of books or the shapers of landscapes now far away, we very rarely play alone, and even the act of sitting quietly with a book in a public space is an act of leadership or following. The playground lays this bare.

-->

### Best-Outcome Economies

One measure of the health of an economy is whether the interests of businesses are aligned with the interests of their customers or users.  To be quantitative about this, we can check whether sales and spending figures match up with long-term satisfaction ratings, with customers' health, and with other measurements of well-being.

The epidemics that currently face the modernized countries--obesity, social isolation, depression, a lack of meaningful work, and so on--indicate just such a misalignment.  We are seeing sales that make customers' lives worse. Industry and marketing focus on clicks, views, downloads, and sales, rather than on our long-term desired outcomes.

<div class="sidenote">
CONSUMER SPENDING. If we imagine an outcome economy--our society re-gearing around longer-term forms of satisfaction: How would it be different from the consumption economy pictured, that we have today? Click for an
<a href="http://www.nytimes.com/interactive/2008/05/03/business/20080403_SPENDING_GRAPHIC.html">explorable map from nytimes.com</a>
</div>

To *outcome-align our society* would require both (a) detecting when and how best our true interests are satisfied, and (b) either reshaping traffic / sales or otherwise changing incentives for businesses and organizations to match those reports of long-term satisfaction.

It is likely that more will be required to reshape demand than just better menus, but the collective wisdom gathered via retrospective reviews can tell us when the incentives are misaligned. We can then decide as a society how to disincentivize businesses from leading us into activities we regret.

The potential switch is from marketplaces that are mostly about transactions to marketplaces for personal outcomes, and means redirecting attention and cashflow towards providers and sites that produce the best outcomes, best livelihoods, and lowest costs in users' lives, rather than those with viral loops, network effects, marketing deals, or other advantages.

## Conclusion

*"We shall once more value ends above means and prefer the good to the useful. We shall honor those who can teach us how to pluck the hour and the day virtuously and well."* -- John Maynard Keynes

In 1930, economist John Maynard Keynes claimed that the sales metrics that drive capitalism would at some point start driving us in the wrong direction.  Perhaps that time has come.  Can we measure demand with the question “what are you glad that you bought?” instead of “what did you buy?”, or online, with “what are you glad you clicked?” rather than “what did you click?”  Can we tether the lifeblood of our economy--cashflow, attention, and resources--to reports of informed and fulfilled lives?  Would people then work together to create the best outcomes for one another--in much the same way they currently collaborate to generate sales, clicks, and views?

## Summary

* A standard for judging the device's responsibility to its owner is advisable.

* At the platform level, devices should give us options aligned with our own interests--rather than in the interests of increasing engagement metrics for media companies, and should present those options in a way that lets us choose well.

* This requires giving us informed, expressive, and social choices.

* Enabling informed choice may require that options listings be augmented with likely time costs, likely money costs, the unlikelihood of various hoped-for outcomes, the likelihood of unhoped-for outcomes, whether there are less-costly or better-success-rate options for similar outcomes, and even whether our hopes themselves are likely to change.

* Enabling expressive choice means that options presented must relate to the values and outcomes that are, or that could be, important to us.

* A public data service about people, choices, and outcomes could change search costs and behavior.

* These models of informed choice, expressive choice, and collective wisdom relate to microeconomics of addiction, behavioral economics, and current social issues like obesity, depression, a lack of meaningful work, sexual dysfunction, clickhole media / the buzzfeedification of media, and "internet addiction".

## Recommended Reading

*To be written.*

## Acknowledgements

Thanks to [Tom Chi](http://tomchi.com) and [Glen Chiacchieri](http://glench.com) for direct feedback on drafts of this paper.

Thanks to [Tristan Harris](http://twitter.com/tristanharris), [Bret Victor](http://worrydream.org), [Jonathan Harris](http://number27.org), [Aza Raskin](http://azarask.in), and [Dan Mosedale](http://github.com/dmose) for motivating and clarifying discussions.

<!--
If 60% of people regret visiting youtube on a tuesday night, you should know that before you click a suggested video.

*"Young people are threatened... by the evil use of advertising techniques that stimulate the natural inclination to avoid hard work by promising the immediate satisfaction of every desire."* -- Pope John Paul II

“Ads sell a great deal more than products. They sell values, images, and concepts of success and worth.”

“It did what all ads are supposed to do: create an anxiety relievable by purchase.”

-->

</div>