Recommender Systems

Thumbs down on Netflix ditching of 5-star ratings for a thumbs-based system

Netflix_logo.svgThe Netflix video streaming and DVD service announced Thursday that it is switching from a 5-star rating system to a simpler thumbs up / thumbs down system. I've been a Netflix user (and fan) for many years, and love their personalized ratings predictions. I have often used their model in presentations and brainstorming involving other services that could benefit from that kind of personalization. I think this change is a bad idea.

I called the Netflix Help Center (866-579-7172), and the customer service representative I spoke with told me they were eager to receive feedback on this topic, especially feedback that specifies why users are in favor or not in favor of the proposed change. I shared several reasons why I thought it was a bad idea and I want to share those reasons here, in the hope it may encourage other Netflix users - especially those who share my view that it's a bad idea - to contact Netflix and provide feedback.

Granularity is Good

The main objection I have to the proposed change is that I make careful distinctions in both the ratings I give to movies I have seen and on the personalized predicted ratings Netflix offers for movies I have not yet seen. I probably watch, on average, 2 hours of TV a week, 1 DVD every month, 1 movie in a theater every 3 months. I hardly ever watch video content on Netflix, YouTube or other streaming sources. So I'm probably an outlier on several dimensions.

That said, on the rare occasions when I do want to watch a movie - at home or in a theater - I will only watch a movie for which the personalized predicted Netflix rating is at least 4.0. Since I know the personalized prediction accuracy is dependent (in part) on my own ratings, I am very careful in how I rate movies. I use the following interpretations for the 5-star scale:

  • 5 stars: a movie I liked so much I've seen it several times and/or would enjoy seeing again
  • 4 stars: a movie I liked a lot, but am not interested in seeing again
  • 3 stars: a movie I liked, but would probably have preferred to spend my time watching something else
  • 2 stars: a movie I didn't like, and probably didn't watch much of
  • 1 star: I don't know if I've ever seen a 1-star movie, and certainly don't want to ever see one

While some people may find it easier to give a thumbs up or thumbs down rating (which I will refer to hereafter as a thumbs-based rating), I would find it more difficult. I envision the following mapping from my 5-star schema to thumbs-based ratings:

  • Thumbs up for a 4-star or 5-star movie
  • No rating for a 3-star movie
  • Thumbs down for a 1-star or 2-star movie

Given that I rarely see a 2-star movie, I would probably only be giving thumbs up ratings in the proposed new scheme, and predict that the lower volume of ratings combined with the lower granularity of ratings would result in less accurate Netflix rating predictions.

Quality vs. Quantity

Speaking of quantity, the Verge article reported that Netflix saw a 200% increase in the number of ratings among the test group who used thumbs up or thumbs down, compared to the number of ratings using the 5-star rating group.

The article doesn't report on the change in the number of users who submit ratings using thumbs up or thumbs down, nor is it clear whether a specific control group was used in the experiment. Based on their marvelously detailed posts in the Netflix tech blog, especially the posts on their recommender systems, I suspect they were very careful in the way the designed the experiment. Perhaps more details will eventually be reported there.

The article also doesn't report on the quality of the recommendations under the thumbs-based rating system. More is not necessarily better, and it is not clear what kind of impact the increased quantity had on the perceived quality of predictions based on the new system.

Emarketer_video_watching_2015Given that the average U.S. adult consumes 5.5 hours of TV, movies, games, and other video content per day, I suspect most users are less discriminating than I am with respect to what they will watch. It may be that the quality of recommendations using the new system serves high-volume - or even average-volume - video consumers as well or better than it would serve low-volume video consumers. But if my supposition that higher volume video consumers are less discriminating is correct, then the increase in quality may not have much impact on the amount consumed. And since Netflix charges flat monthly rates, those of us who consumer relatively little video content are paying just as much as those who consume large amounts of video content .. and if the recommendation quality declines for someone like me, who consumes little content, and the quantity of video I consume similarly declines, I am more likely to discontinue the service than a high-volume consumer who might consume less if the quality of recommendations is not as good (due to fewer ratings). But if they are already consuming a large quantity of video, I don't understand what problem is Netflix trying to address.

Returns on Investments

Spotify-vs-pandora-whats-the-best-app-for-streaming-musicThe article draws an analogy between Netflix ratings and Spotify thumbs-based ratings, which I think is an inappropriate comparison point. I use both the Spotify and Pandora streaming music services (in fact, I'm a paid subscriber for both (I hate commercials in any medium)), but rating a song that lasts a few minutes is very different - in my view - from rating a movie that lasts a few hours. I'm much more willing to provide a finer granularity rating (e.g., on a 5-star scale) for an experience that will last hours vs. minutes.

I think a better comparison point would be Yelp, which uses 5-star ratings for restaurants and other service providers. I'm willing to provide ratings on a 5-star scale for restaurants, because it represents a more significant investment of time (and money). I would even consider TripAdvisor, an online service for reviews and ratings of hotels and other destinations and activities associated with traveling, a better comparison point than Spotify, as pl

Personalized Ratings for All

In fact, I think both Yelp and TripAdvisor could benefit from adopting the potentially-soon-to-be-former Netflix personalized rating scheme. I am growing weary of wading through reviews of restaurants on Yelp from people who rant about the bartender not paying attention to them, or a special event dinner that went awry, or from anyone who doesn't share similar tastes in restaurants to me. I would love it if Yelp would offer a personalized rating, or at least let me read reviews from people like me.

TripAdvisor ratings have become almost useless to me. It appears that many hotels are carpet-bombing guests with email invitations to review their stay, and the result seems to be that many places now have an overwhelming abundance of reviews from people who have only posted one review. I consider most newbie reviews nearly useless, both because they tend to be short and uninformative, and because there is no way to know what kind of other places the person has reviewed, so I can't tell how much the reviewer is like me.

I could rant further on the decline of both of these services - which I once found far more useful - but I will let it go (for now). I wanted to compose this post because throughout all the years I've been a Netflix user, the service has only gotten better (as I gave it more ratings upon which to make recommendations), and I'd hate to see yet another beloved rating, review and recommender service decline.

If you feel similarly, I urge you to call Netflix soon, as they are reportedly planning to roll out the new thumbs-based rating system in April.

musica requitur sequitur: songs that must be played in sequence

Pandora_logoSince my shoulder surgery 4 weeks ago, I've been spending a lot of time developing software and listening to Pandora. The pain meds (oxycodone & hydrocodone) put me in a bit of a brain fog, limiting the effective breadth and depth of my thinking (and doing), but a reasonably well-defined coding task seemed ideally suited to my power of concentration ... and I've always found listening to music while coding helps put & keep me in "the zone".

The Pandora fremium online music service has developed such an accurate model of my preferences over the past few weeks that I've upgraded to Pandora One. The annual subscription version of the service has eliminated commercials and increased the length of time I can listen per day, and eliminates the pause and prompt asking "Are you still listening?" if I don't interact regularly with the site.

The one annoyance that remains is that there are certain songs that I believe should never be played without also playing the song that immediately follows them on the original album / CD. This strikes me as the musical opposite of a non sequitur - examples of which are virally proliferating as we slog through the U.S. presidential election season - so I propose the following name for this phenomenon:

musica requitur sequitur

DarkSideOfTheMoonAmong the examples I've noted over the past week [update: now augmented with examples shared in comments]:

  • Sgt. Pepper's Lonely Hearts Club Band / With A Little Help From My Friends (Beatles, Sgt. Pepper's Lonely Hearts Club Band, 1967)
  • Falling In and Out of Love / Amie (Pure Prairie League, Bustin' Out, 1972)
  • On the Run / Breathe / Great Gig in the Sky (Pink Floyd, Dark Side of the Moon, 1973)
  • Any Colour You Like / Brain Damage / Eclipse (Pink Floyd, Dark Side of the Moon)
  • Happiest Days of Our Lives / Another Brick in the Wall Part 2 (Pink Floyd, The Wall, 1979; contributed by Eric)

I'm probably dating myself with these examples, and by my uncertainty about whether contemporary bands are producing songs that are intended to flow so naturally from one to the other. Perhaps it was solely or primarily a trend of the late 60s and early 70s.

I'll update the list with additional examples as I encounter them, and would welcome any other examples anyone is inclined to share in the comments.

Recommender Systems in Retail Stores: Bridging the Gaps between eCommerce and Physical Shopping

Advertising Age posted an article yesterday on Olay Translates Killer Online App to Retail Aisles, describing some recent trial deployments of special kiosks in physical stores that give shoppers access to online recommender systems. These systems have access to online representations of offline inventories - or, at least, product lines typically carried - in the stores, and help bring some of the advantages of electronic commerce (e.g., more detailed information and personalized recommendations on products) to bricks and mortar stores.

Shoppers - online and offline - often suffer from the paradox of choice: a range of options so vast that it can feel overwhelming (e.g., the "250 varieties of cookies, 75 iced teas, 230 soups, 175 salad dressings, 275 cereals and 40 toothpastes" that Barry Schwartz mentions in his TED presentation on the topic) and result in low customer satisfaction, regardless of which option is selected. Recommender systems help people navigate broad ranges of options in the digital world, but without the assistance of a salesperson, there has been little help for shoppers in physical stores ... until recently.

For the growing number of consumers who prefer the online experience to traditional shopping, the ease of finding products and getting recommendations clearly is a draw, said Carter Cast. Mr. Cast, a former CEO of and head of strategy for Wal-Mart Stores in the U.S., became CEO of fledgling specialty online retailer Netshops late last year.

Because of expectations created by web shopping, consumers increasingly expect offline stores to have the goods they want and make them easy to find, Mr. Cast said. "So the ante is raised in the physical world."

In an effort to meet these expectations, Procter & Gamble has developed and deployed a version of their popular Olay For You online recommender system (tag line: "a little about Olay, a lot about you") that bridges the gap between the wealth of online personalized information and the experience of customers in offline WalMart stores. In either case - visiting the web site or at a WalMart kiosk - users characterize their general wants and needs, selecting from among options such as

  • I want to see a visible improvement in my skin
  • I'm happy with my skin but want to help it be the best it can be
  • I want to look good for a special occasion
  • I want to keep up with the latest skin care products

and then progressively reveal more specifics about themselves (e.g., their age, their skin type and color, whether they are experiencing hormonal changes, and other lifestyle issues such as "not enough 'me' time"), until the system recommends a skin care product that is deemed likely to be right for them. Screenshots of my profile and recommended Olay products can be seen below.



The strictly online version offers the capability of optionally remembering the user's profile (associated with their email address); it's not clear from the article whether or how the kiosk version, designed by Talk Me Into It ("GPS for the overwhelmed buyer") allows this, nor whether it allows offline shoppers to access their previously created online profiles at the kiosks, nor even whether the system has real-time access to the store's inventory. And, of course, it remains to be seen how willing customers will be to reveal some of the personal details the online recommender system asks when they are interacting with the system in a public setting like a retail store aisle.

Another system mentioned in the article is the Search Engine in the StoreTM developed by Evincii. The description of the system on Evincii's web site articulates a rather comprehensive value roposition:

Evincii's in-store search engine recommends precise and relevant products to consumers in brick-and-mortar and online stores. Our network delivers an interactive, targeted, search-based platform at the point of decision with proven, category-wide and brand-specific sales lift. Shoppers receive personalized advice in seconds. Retailers get happy customers, improved sales efficiency, and increased margins. Product manufacturers get the opportunity to present their products to individual shoppers just before the point of purchase.

EvinciipharmassistadageAccording to the article, one Evincii kiosk system, which I believe is [cleverly] named "PharmAssist", has been deployed in Longs Drugs stores in California (Evincii is headquartered in Mountain View) since 2006, and includes targeted advertising as part of their search capability:

Johnson & Johnson is an initial advertiser on the system, which allows advertisers to place ads similar to online display ads, including video, around search results.

But like Google or other search engines, Evincii looks to return "organic" results only based on the criteria shoppers input, such as their symptoms, said Charles Koo, CEO of the private-equity-backed venture. Then, once they've selected a product, the kiosk helps them locate it on the shelf.

I don't know anything about the full range of advertising available for inclusion with "organic" search results, but I find myself musing about how Longs' customers might respond to video advertisements for  "sensitive" products such as Trojan condoms or Ex-lax ... especially if the advertisements include an audio component. The AdAge article also expresses some skepticism.

Mr. Koo, however, said Evincii's research at Longs indicates that 15% to 18% of visitors to OTC-drug departments use the kiosks, numbers similar to those that ComScore found last year of consumers who use online search to research package goods. Stores using the kiosks, he said, had category sales lifts of 3% to 6%.

My observations and judgments are markedly different from those shared by Evincii. Although I don't recall visiting a Longs Drugs store in California during the last two years, I'm reminded of how annoying I found a kiosk deployed at another drug store - I think it was a Walgreens - on El Camino Real in Palo Alto. Whenever anyone got near, the kiosk would loudly offer "Can I help you find something?". While I was there waiting for a prescription to be filled, about a year ago, I informally observed the kiosk for about 15 minutes - from a safe distance - and while I heard the audio offer of assistance at least a dozen times, no one stopped to take advantage of the offer and interact with the kiosk ... and I wondered how many other customers, like me, avoided that section of the aisle like the plague.

Despite having some reservations about some of the examples reported in the AdAge article, I do think that bridging the gaps between online [commerce] and offline [shopping] holds great promise in general. The art - and science - is to design the bridges in a way that offers compelling value to all stakeholders, and to situate them in the kinds of spaces where that value can best be realized.

[Update, 2008-06-05]

I received an email from a reader with more information about the OlayForYou system and the kiosks deployed in stores. The reader would prefer not to be identified, but is willing to permit me to share the information from the email:

  • The OlayForYou system does not track behavior over time, a key feature of many recommender systems.
  • The system was designed around the concept of "buy soon" - rather than "buy now" -  a shopping list you could cross off one item at a time as your budget allowed.
  • The kiosks used Rivet Digital Touch Screens and were deployed in WalMart stores. The reader was not sure where, how many or whether they are still in use. The kiosks do not have network connectivity, and so have no access to personal profiles or real-time inventory.
  • The kiosk questions are simpler than those on the web site, due to time and privacy constraints of in-store use.

Many thanks to the reader who offered the additional information!

Innovating at MyStrands, Seattle

It's been a little over a month since I left Nokia and started principally instigating at MyStrands, Seattle. Most of my time thus far has been devoted to talking with people and looking at places, as my top two initial instigative goals are to attract a dream team and setup shop in a dream space. Eventually, we'll make progress on other p's - prototypes, papers and patents - but not without the right people and only (or, at least, more easily) in the right place.

I'm making some progress on these initial goals - I will be making official announcements when formal transitions take place - but meanwhile I thought it would be helpful (to me, at least) to write a little bit about what kinds of innovation - and what kinds of innovators - I hope to facilitate with this great new team and great new space! Taking a cue from one of Glenn Kelman's pearls of wisdom - "We never say 'I'" - during an inspiring presentation at my crash course in entrepreneurship (NWEN's Entrepreneur University 2005), I'll be using "we" liberally below, even though "we" is, technically speaking, "I" at this particular moment.

The mission of the Seattle Innovation Team for MyStrands is

To design, develop and deploy technologies that weave together the various strands of our activities, interests and passions to bridge the gaps between the digital and physical worlds and help people relate to the other people, places and things around them in ways that offer value to all participants.

That's quite a mouthful (even for me) so I want to unpack that a little:

  • weave together the various strands of our activities, interests and passions: MyStrands started out as MusicStrands, a web application that can recommend new music based on the music you listen to ("what you play counts!"). Since then, the company has branched out into other types of media (e.g., MyStrands.TV), and we want to further extend this extension to additional types of media, as well as other digital representations of our activities, interests and passions.
  • bridge the gaps between the digital and physical worlds: with the growing wealth of digital representations of our activities, interests and passions, and the proliferation of mobile devices and wireless connectivity, there are increasing opportunities to create new value by opening portals to that wealth in the physical world, either through mobile social computing (MoSoSo) applications or more situated social computing (SiSoSo) applications, such as proactive displays.
  • help people relate to the other people, places and things around them: we all long to feel a sense of belonging and connection to other people we encounter, the places we inhabit and the things we see (or at least some of those people, places and things); our technologies will be designed to help real world communities better enjoy the benefits of virtual communities, digital communications and electronic commerce.
  • offer value to all participants: one of the things I learned - the hard way - during my earlier entrepreneurial endeavor (Interrelativity, Inc.) was the importance of aligning innovative social technologies with viable business models; although our primary focus will be on technical innovations with significant - and positive - social impact, we want to do so in an economically sustainable way that enriches all stakeholders.

As for the types of innovators (we don't call ourselves researchers - or developers - at MyStrands, though we will be doing both) we hope to attract, the primary criteria will be a passionate commitment to the mission of the lab. Of course, following the precepts of Joel Sposky, we also generally want smart people who can get things done. Among the more specific types of smarts that we value are insights into and experience with social computing, mobile and ubiquitous computing, human-computer interaction, many flavors of design (user interface design, interaction design, user experience design, visual communication design), web programming, rapid prototyping, personal and social psychology, economics, business models ... and, of course, recommender systems.

I'm planning to follow the lead of Lars Erik Holmquist with respect to [one of] his goals for the PLAY Research Group: build a multidisciplinary team composed of multidisciplinary people. As Anne Galloway relayed this idea (in what may be an unexpected recursion based on something I may have sent / said to her about Lars Erik's CHI 2000 organizational overview talk):

Alan Kay once remarked that he was attracted to the MIT Media Lab because of the..."attempt to collide technology with the arts, rather than [to] collide technologists with artists," and continued "You're always better getting people who have already had that collision in themselves." In PLAY, rather than composing a multi-disciplinary group, we try to have a group of multi-disciplinary people ... No group member specializes in only one topic. A typical member has a degree in a relevant field such as computer science, informatics or fine arts, but a strong interest in several other fields such as electrical engineering, linguistics, literature, film, or music. Whether accompanied by academic degrees or not, a wide range of interests is seen as a vital factor in the composition of the group.

Ideally, we will compose a diverse group of diverse people, with a variety of skills, from a variety of backgrounds, who respect each other and work well together, even though - or perhaps because - we may not always agree with each other (indeed, I hope we won't always agree with each other). I have enjoyed many conversations with many talented people so far, and I welcome the opportunity to initiate or renew conversations with other talented people. If current trends continue, we may be able to assemble a dream team without ever having to compose or post a formal job description.

As I mentioned in my review of First Break All the Rules: What the World's Greatest Managers Do Differently, I subscribe to the philosophy of Bud Grant, former head coach of the Minnesota Vikings:

You can’t draw up plays and then just plug your players in. No matter how well you have designed your play book, it’s useless if you don’t know which plays your players can run. When I draw up my play book, I always go from the players to the play.

Given my commitment to multidisciplinarity, I'm going one step further and not even specifying player positions at this early stage, hoping that we will be able attract just the right people with just the right talents to accomplish our innovation mission.

[Update: I forgot to mention that we have a number of more formally specified job openings at many of our other sites around the world. The (Senior) Mobile User Experience / Interaction Designer position may be of particular interest to some of my international friends.]

Entrepreneurial Karma: Call for Early-Stage Recommender System Startups

This is pretty cool (though I'm admittedly biased): a mid-stage startup (MyStrands, the company I work for) that has recently secured funding now offering an opportunity to fund an earlier stage startup - a sort of entrepreneurial karma, where we keep the investment flowing in ways that will [hopefully] benefit us all. This initiative is inspired, in part, by the Y-Combinator, but is more narrowly focused (recommender systems). [BTW, one of the partners in Y-Combinator, Paul Graham, writes provocative essays that I highly recommend to anyone interested in entrepreneurship.]

Here are the details (from the MyStrands blog announcement, Looking for the best early-stage recommender start-ups):


MyStrands is announcing today the “Strands $100,000 Call for Recommender Start-ups“.

We seek to identify the best early-stage project in the area of recommendation technologies, considering the technology, business opportunity and team behind the project (without limitations as to which field the technology is applied).

The Winner will be offered an investment of $100,000 from Strands, Inc. the parent company of MyStrands.

Candidates should submit a one-slide presentation in quad-chart format (example, more examples) by September 15th, 2008 to, together with the team bios (in addition to this, an optional 2-minute video uploaded to YouTube describing the start-up enterprise would be highly appreciated).

Eligibility: The Call is open to individuals or sole proprietors and privately held businesses throughout the world.

Five Finalists will be invited to present their projects during the ACM Conference on Recommender Systems (RecSys08) next October 23rd to 25th, 2008 in Lausanne, Switzerland. Finalists will be announced on October 6th. All Proposals will be judged using the following judging criteria: (a) implementation and integration of recommendation technologies, (b) originality and creativity, (c) likelihood of long-term success and scalability, (d) effectiveness in addressing a need in the marketplace, and (e) team bios.

Five grants. Each Finalist will obtain a $1,500 travel grant to attend RecSys08; Strands, Inc. will also cover the registration fees for the Conference, for one person per Finalist.

The final selection process will include on-site presentations of each project during RecSys08. Finalists will make three presentations of 5 minutes each (focused on technology, business and the team respectively) in front of the Jury and the attendees of the Conference.

The Jury will be composed of renowned experts in the academic, industry and venture capital communities.

The Winner will be announced on October 25th, 2008 during the Gala Dinner at RecSys08. The Winner will receive a commemorative plaque and an offer of a $100,000 investment in the form of a convertible loan.

Proposal submission period begins on March 12th and ends on September 15th, 2008.

Further information: Please visit or contact the Strands $100,000 Call for Recommender Start-ups organizers at

Important dates:
March 12th: Proposal submission period begins
September 15th: Proposal submission period ends
October 6th: Five Finalists are announced
October 23rd-25th: Presentations of Finalists at RecSys08
October 25th: Winner Announced

Contact and updates:
Email proposal to:

The Paradox of Choice: Decisions, Happiness and Appreciation

In addition to seeding my last post - on Dark Nights of the Soul - by sending me a link to an evocative image, Yogi also sent me a link to a 20-minute video of Barry Schwartz giving a presentation on The Paradox of Choice a few years ago at a TED conference.

Paradoxofchoice The presentation is derived from Barry's book, The Paradox of Choice: Why Less is More, about which he says (during the video presentation), "I wrote a whole book to try to explain this [paradox] to myself" ... reminding me of why I blog ... or why David Whyte writes poetry ("Poetry is the art of hearing yourself say things you didn't know you knew" - (well, at least I think these are all related)).

Barry offers an engaging theory on why increasing choices often makes people miserable:

  1. Regret and anticipated regret
  2. Opportunity costs
  3. Escalation of expectations
  4. Self-blame

He is not railing against choice(s), but instead arguing that the "official dogma" - having more choices leads to more freedom and [thus] greater welfare - is wrong, that we pass a point of diminishing returns after which welfare - or happiness - decreases as the range of choices increase. As he says "some choice is better than no choice, but more choice is not necessarily better than some choice". While I generally and enthusiastically agree with many of his points, I think we differ on where the points of diminishing returns start, whether or how to set boundaries near those points, and whether there are other ways of approaching choice that may affect these points.

Barry draws some of of his examples of overwhelming choices from his supermarket, which stocks 250 varieties of cookies, 75 iced teas, 230 soups, 175 salad dressings, 275 cereals and 40 toothpastes (several years ago, a Washington Post article on Toothpaste Proliferation Syndrome reported finding 179 varieties of toothpaste in a virtual stroll down the aisle at

One of the side effects of so much choice is that it increases the likelihood that we'll make a "wrong" choice, and the corresponding likelihood that we'll regret our choice ... and the likelihood that we'll anticipate regretting our choice:

Even if we manage to overcome the paralysis and make a choice, we end up less satisfied with the result of the choice than if we had fewer options to choose from.
With a lot of different salad dressings to choose from, if you buy one and it's not perfect ... it's easy to imagine you could have made a different choice that would have been better. And what happens is this imagined alternative induces you to regret the decision you made, and this regret subtracts from the satisfaction you get out of the decision you made even if it was a good decision. The more options there are, the easier it is to regret anything at all about the option that you chose.

TurtlesallthewaydownDecision_tree_modelFurther on in the presentation, Barry distinguishes between time spent making decisions vs. acting on those decisions, which evoked an image of spending more time computing in the nodes vs. traversing the branches of a decision tree. But upon further reflection, this representation of activity led me to start wondering what proportion of of the time we are acting on previous, or higher level, decisions, is actually spent making lower level decisions, many of which we may not even be conscious of ... whether, in effect, it's decisions all the way down.

Shifting from decision theory to political theory, representative democracy strikes me as one model for modulating choice, wherein we elect representatives who we then hope will make good decisions regarding a large number of options across a wide range of topics on our behalf (behalves?). I rarely hear people complain about the overwhelming range of choices in political candidates (in this country), although I often hear people complain about decisions made by our "elected" politicians ... especially during a U.S. presidential election year. 

I'm reminded of much I've heard and read about "tastemakers" and "trendsetters", and I imagine we often tend to gravitate toward official and unofficial "authorities" to reduce our anxiety over the array of choices. One of the benefits of this tendency is that it can simplify our lives. However, the shadow side of this tendency reflects Don Miguel Ruiz' notion of "domestication" (described in the introduction of his book, The Four Agreements), in which we learn to defer to authorities as children, and eventually learn to constrain ourselves based on what we think the authorities would want us to think, feel and do. [The Four Agreements are about unlearning this domestication, or at least consciously making agreements about what rules - and rulers - we are willing to follow. I've written about two of them before - don't take anything personally and always do your best - and will surely find a pretext to blog about the other two - be impeccable with your word and don't make assumptions.]

This delegation of power to authorities is what bothered me most about the presentation. Barry complains that doctors no longer tell us what to do, but instead list alternatives we might consider. Personally, I like doctors presenting alternatives with relative benefits and risks vs. telling us (me) what to do; the latter, more "traditional" (and "authoritative") approach strikes me as disempowering, and I would greatly prefer to deal directly with the misery entailed by being offered a multitude of health care alternatives than to suffer the degradation of condescending "doctor's orders".

The idea of delegation is closely related to surrogation, which reminds me of Dan Gilbert's ideas on how we make decisions (poorly) and how we ought to make decisions (using a surrogate). As I'd noted in an earlier post on Dan's book, Stumbling on Happiness, he argues that

rather than relying solely on our selves (and our fallible memories) to imagine how happy we will feel in some future state, we should capitalize on the experience of others by inquiring about the happiness of those who are already in the future state we are considering ... the problem [then] is figuring out which others we ought to consult in estimating our future ... I want to know what people like me like.

[I'll also note, here, that I recently discovered a TED video of Dan's presentation on Stumbling on Happiness].

Given my renewed research into recommender systems, and some recent ruminations about re-rethinking recommendation engines, I see how such systems can also play a key role in effectively addressing the increasing array of choices we face in our lives ... and in helping me find people like me ... and what those people like.

What people [like me] like reminds of David Whyte's perspective on why we are liked (or loved), expressed through his poem, "This Time":

Those stars told him
they loved him only
for what he loved himself.
They did not love him
for who he was.

So, people like me [might] like me for what I like.

Toward the end of his presentation, Barry wryly comments "the secret to happiness is low expectations". I've been a student of happiness for some time, and am intrigued with many dimensions of the art, science and business of happiness. And so, I again turn from the science to art - from psychology to poetry - and invoke Oriah Mountain Dreamer's perspective on the secret to happiness ... which not only offers a contrast to Barry's, but also seems to conflict with the sentiment expressed by David Whyte (which is particularly incongruent, for me, as one of his workshops had inspired her earlier prose poem, The Invitation, and I find a great deal of congruence between that poem and his poem, Self Portrait). Oriah expresses a view which has more to do with greater appreciation - and less to do with lower expectations - and seems to admit the possibility that we might well love ourselves for who we are (not just for what we love) in the Prelude to her book, The Dance:

What if your contribution to the world and the fulfillment of your own happiness is not dependent upon discovering a better method of prayer or technique of meditation, not dependent upon reading the right book or attending the right seminar, but upon really seeing and deeply appreciating yourself and the world as they are right now?

So maybe what we really need in the next generation of recommender systems - as a way out of the paradox of choice - is new mechanisms to help us better appreciate ourselves, and the people, places and things around us ... and perhaps new ways for expressing our love for people, places and things.

In any case, I think a healthy dose of Susan Jeffers' "no-lose decision model" (from her book, Feel the Fear and Do It Anyway ) offers a remedy for avoiding the misery and regret that Barry talks about. I've mentioned it before, but I'll include it yet again, as I think it is relevant (and helpful):

Before you make a decision:

  1. Focus immediately on the no-lose model (whichever path you choose will provide learning opportunities … even if it’s learning what you don’t like)
  2. Do your homework (talk to as many people as will listen … both to help clarify your own intention and to get alternative perspectives)
  3. Establish your priorities (which pathway is more in line with your overall goals in life – at the present time)
  4. Trust your impulses (your body gives you good clues about which way to go)
  5. Lighten up (it really doesn’t matter – it’s all part of a lifelong learning process)

After making a decision:

  1. Throw away the picture (if you focus on what you expected, you may miss the unexpected opportunities that arise along the new path you’ve chosen)
  2. Accept total responsibility for your decision (don’t give away your power)
  3. Don’t protect, correct (commit yourself to any decision you make and give it all you got … but if it doesn’t work out, change it!)

Although we may not want to apply the full range of this model in every choice we make - talking to as many people as will listen about which toothpaste to buy seems a bit extreme - but lightening up and letting go seem like good practices to apply to all our choices.

Re-rethinking Recommendation Engines: Psychology and the Influence of False Negatives

Alex Iskold posted an interesting article on Rethinking Recommendation Engines on ReadWriteWeb yesterday. I like (and recommend) his crisp and clear delineation of different types or sources of recommendations - personalized (based on your past behavior), social (based on past behavior of others who are similar to you) and item-based (based on the recommendable items themselves) - and his emphasis on the importance of incorporating psychological principles, not just technological ones, into the design of effective recommendation engines. [I also like (and recommend) Rick MacManus' associated recommendations on 10 Recommended Recommendation Engines, but that may be biased by MyStrands' prominent placement in that list.] However, I take issue with - or at least re-rethink - some of Alex' contentions regarding the road to successful recommender systems being paved with false negatives.

First, I want to agree with Alex (and Gavin Potter, the Guy in the Garage that Alex references) about the importance of psychology in technology design and in general ("Enhancing formulas with a bit of human psychology is a really good idea") and the value of recognizing and capitalizing on human inertia. However, his characterization of inertia - the tendency of our ratings to be heavily influenced (or primed) by other recent ratings - seems more characteristic of a primacy or recency effect than inertia (as I understand these concepts). However, I do think that inertia plays an important role in the adoption and use (or non-adoption / non-use) of any technology - people do not tend to change much or even expend much effort, unless or until sufficient incentive is provided.

So I think the inertia problem, with respect to recommendation engines, is more one of motivating users to rate things ... and I actually think the Netflix ratings system for movies (which provides the basis for much of the article) is an outstanding example - it doesn't require much effort (you are automatically prompted for a rating whenever you login to the site after having sent a DVD back), and the more you rate, the better the recommendations you receive, offering intrinsic vs. extrinsic motivation ... and explaining why the system has motivated millions of its users to contribute an estimated 2 billion ratings. [Aside: I see that ReadWriteWeb is offering an extrinsic incentive for comments and trackbacks - a chance to win an Amazon gift certificate - but I was already planning on adding a trackback for intrinsic reasons.] In any case, however one labels these psychological influences - inertia, priming and/or recency - they are important to incorporate into the design of recommendation engines, and the systems that use them.

Further along in the article, Alex distinguishes false positives - recommendations for things that (it later turns out) we do not like - from false negatives - recommendations against things (it would later or perhaps likely turn out) we do like, and correctly recommends leveraging false negatives more effectively in the design of recommendation engines. [And just to round things out, in case it isn't obvious, true positives are recommendations for things that we will / do like, and true negatives are recommendations for things that we will / do not like ... and thanks to Eric for helping me set the record straight with respect to "do likes" and "don't' likes" in my description of false negatives (!)]

Unfortunately, he extends this thread to some propositions that lie beyond my comfort zone:

We do not need recommendations, because we are already over subscribed. We need noise filters. An algorithm that says: 'hey, you are definitely not going to like that' and hide it. ... If the machines can do the work of aggressively throwing information out for us, then we can deal with the rest on our own.

Now, on the one hand, I am sympathetic to the problem of information overload. However, as I noted in my notes from CSCW 2006, Paul Dourish pointed out that this is not a new problem:

One of the diseases of this age is the multiplicity of books; they doth so overcharge the world that it is not able to digest the abundance of idle matter that is every day hatched and brought forth into the world.

-- Barnaby Rich (1580-1617), writing in 1613 (!); quoted by de Solla Price in his 1963 book "Little Science, Big Science."

I'm also reminded of James Carse's observation about evil in his marvelous (and highly recommended) book, Finite and Infinite Games:

Evil is never intended as evil. Indeed, the contradiction inherent in all evil is that it originates in the desire to eliminate evil. "The only good Indian is a dead Indian."

I think that too aggressively filtering out [presumed] false negatives can render us more easily manipulated by technology ... and the people and organizations who control technology. Although there is considerable debate about what Web 2.0 is, one of its key ingredients is surely the provisioning of architectures of participation, in contrast to the "command and control" paradigm of earlier technologies (and eras). One of the beneficial side effects of the growth of Web 2.0 - for me - has been enhanced opportunities for serendipity, and allowing more false negatives is likely to yield fewer instances of serendipity. Furthermore, I believe increasing the probability - or acceptability - of false negatives may have the unfortunate consequence of moving further up the head of the long tail ... and/or further down toward the lowest common denominator(s). Book burning lies at or near the extreme end of the "acceptance of false negatives" spectrum, though I do not mean to imply that any of these consequences are intended or desired by the article or author.

In earlier chapters of my career, when I was more focused on natural language processing and automatic speech recognition, I became familiar with the concept of Equal Error Rate (EER), which represents a way of measuring the balance between false positives (which yields what is called the False Acceptance Rate, or FAR) and false negatives (False Rejection Rate, or FRR). The documentation for the BioID biometrics system SDK from HumanScan provides a nice articulation of these concepts, including the graph below:


Perhaps the solution to the tension between false positives and false negatives in recommender systems is to incorporate some kind of control for the user to specify an acceptable balance or threshold (which may default to the EER)  ... although that would also require devising a solution to the tension between user inertia and input ... but that simply provides additional corroboration for Alex's primary argument that we need to incorporate more psychology into our designs of good - or better - recommender system technologies.