Variants of the golden rule

There are different versions of basic ethical principles that are common across cultures, and may have roots in the evolutionary advantage of cooperation.

The famous New Testament version is “do unto others as you would have done unto you.” A Talmudic version has a subtle, but notable difference, “what is hateful to you, do not do unto others.” The difference may relate to differing attitudes toward proselytizing. Someone who would want to have been converted to Christianity would offer the same benefit to others. By contrast, someone who would prefer not to be proselytized would recommend against inflicting others with one’s beliefs.

In many situations, I like the Wiccan version of the golden rule, “do as ye will an ye harm none.” The cautionary clarification is that “harm none” includes the self – so practices that are pleasurable but self-destructive would be discouraged.

There’s another variant that I wish existed but haven’t seen anywhere. “Do unto others as they would have done unto them.” This assumes that what the other person wants may be different from what you want, and encourages you to treat them as they would wish to be treated. The other ones are easier, the roots of empathy and avoiding harm are found in one’s own feelings – this one requires reflecting on how another’s wants are different.

Wikipedia has a whole catalog of cross-cultural variants, here. What do you think?

Social recommendations and the Eliza Effect

In a post on on Algorithmic Authority one of Adrian Chan’s key points is to displace the critique of the authority claim from the recommendation itself to the users’ acceptance or rejection of the recommendation. “Authority, in short, depends perhaps on the user, not on the algorithm, for it is only on the basis of the user’s acceptance that authority is realized. It is subjectively interpreted, not objectively held.” However, there are a number of problems in severing the communication from its reception.

The example at hand came from Facebook’s flawed friend recommendations. These suggest that you friend or re-contact people, apparently based on an analysis of your social network and communication patterns. These recommendations are particularly annoying because of visual design, and even more so because of social design. The recommendations are featured prominently in the interface, and impossible to suppress without effortful, power-user techniques like Greasemonkey scripts. The big problem is social design. Dennis Crowley, co-founder of Dodgeball and later Foursquare, described the classic flaw with social network based recommendation as the “ex-girlfriend” problem – when an algorithm detects a person that you haven’t communicated with in a while, there is often some very good reason for the change in communication pattern.

The flaw in social network based friend recommendations is related to Adrian’s recent critique of social network analysis. The social network map is not the territory – the visualization of lines of connection based on explicit communications leaves out critical information about the directionality and the content of the communication. A gap in communication may be an unintentional lapse in attention or a rupture; frequent communication may be a sign of closeness or flamewar.

One problem with this misreading of social network analysis results is that the more personalized and personal recommendations, the more likely they are to trigger the Eliza effect, which is “the tendency to unconsciously assume computer behaviors are analogous to human behaviors.” The more a computer impersonates human, the more people will tend to anthropomorphize the computer, and have a strong emotional response to that computer which acts as a human. The converse reason for a strong emotional response to a poorly personalized recommendation can also come into play. The “Uncanny valley” is the name of the disconcerting effect when computer simulations that are nearly but not quite human. People find simulations that are close to human much more annoying than simulations that are more cartoonlike.

It is risky to simply dismiss the effect of pervasive messages, even messages that are not acted upon. Marketers have long considered the psychological effects of communication; marketing messages and frames affect consciouslness even if the listener does not take immediate action, even if the listener superficially ignores the message, even if if the listener superficially disagrees.

You can’t unsee and you can’t unhear. This effect is most visible at the extremes; thus the disturbing effect of chatroulette, the random-navigation chat program that has attracted people in search of random conversation and entertainment, and plenty of flashers. If someoene doesn’t want to see the private parts of random people they should stay off chatroullette; clicking past the flasher doesn’t solve the problem, because you can’t unsee.

Sure, bad social system recommendations are merely annoying; they don’t make us take any action we don’t want to do; but just because we haven’t taken action doesn’t mean the recommendations have had no effect.

The holy grail of internet marketing has been to make recommendations that are powerful and compelling because they are personal, based on a wealth of information based on the user’s personal behavior and actual social network. The lesson for social designers is that it is possible to make recommendations that are not-quite-right, that are more annoying to users because they are more personal. Being personal can be touchy; requiring care and caution, and avoiding overconfidence.

Adrian’s post on Algorithmic Authority has a broader scope, dealing with the larger sociological implications of the idea of algorithmic authority proposed by Clay Shirky, and refining some distinctions on the topic I proposed here. If you haven’t read it, it’s worth consideration.

Buena Vista Social Club and Calle 54

This past weekend I watched the Buena Vista Social Club movie. I had never seen it before for no good reason. I loved the album when it first came out, had it on repeat(n) for months. The Amazon reviews for Buena Vista also referred to Calle 54, a film about Afro-Cuban jazz directed by Spanish directory Fernando Prueba (La Belle Epoque), who is a big fan of Latin jazz. Many of the Amazon reviewers liked Calle 54 better. I thought that judgement was unfair – they were quite different films.

In Buena Vista, director Wim Wenders shows the shabby beauty of late-90s Havana, and the joy and skill of the musicians’ performing and interacting. The movie cuts among travelogue scenes; stories of the musicians in their 70s-90s who were stars in the pre-revolutionary “son” style, returning from obscure retirement, and snippets of music footage. The highlight of the film is a blazing performance of ”El Cuarto de Tula” led by singer Ibrahim Ferrer.

The story telling is implicit and simple on the surface. On the one hand, translating into words cheapens the effect (triumph of the human spirit, universal language, ageless zest for life). On the other hand, the musicians are portrayed as characters in a fable. It’s the way they are interviewed and tell their stories, “I was born in poverty, in a mountain town.” It’s in the way their material and social circumstances are portrayed – jazz pianist Ruben Gonzalez no longer owns a piano and plays at a gym for competitive gymnastics hopefuls. Singer Ibrahim Ferrer continues the folk spiritual tradition as he gives daily librations of rum to his santeria altar; the image is a carving given to him by his mother. A PBS “making of” essay has an interesting perspective about the fable-like quality of the musicians’ stories – as entertainers and performers of folk-derived popular music, they contributed to the mythical flavor of their own stories.

The language and class barriers make a difference. You can see it in the way the musicians talk about Ry Cooder. Ferrer is surprised that a song he tossed off as a warmup was recorded and used on the album – “Ry Cooder liked it.” Throughout his career, Ry Cooder has searched for great music as a student and seeker, and collaborated with musicians from a variety of traditions in various parts of the world. He endured a lot of hassle from the US government in making this film, and tries pretty hard to stay in the background in Buena Vista, instead of taking center stage, white-guy-hero style. Part of the reason the Buena Vista album is great is Ry Cooder’s musical sensibility. Either I am a philistine vulnerable to his accessible cross-cultural raidings, or Ry Cooder has a great ear for affecting music, perfect songs, and clear, unsentimental production. I’m not going to dismiss the movie on political correctness grounds. Still, the economic and political situation shows in the relationships; Cooder is obviously the person giving the Cuban musicians the opportunity to play again and to travel.

The movie has very little overt politics (at the end of the movie, it shows revolutionary slogans on the walls, which have clearly failed to deliver). This is a strength and a weakness; you know that some combination of US and Cuban government activity has contributed to the musicians’ hardships, but you don’t know what or how. The portrayal of Havana has a faded romanticism; which, to be fair, isn’t distinctively colonialist on Wim Wenders’ part, he applies his romantic view of landscape equally to European cities viewed from the eyes of strangers (Wings of Desire) and dusty, declining American rural towns (Paris, Texas).

Unlike Buena Vista Social Club, which highlights the music but places the characters, landscape and story ahead of the music, Calle 54 shows complete, extended musical performances. This makes it less of a filmic work of art, but allows viewers and listeners to get more of the music and the musicians.

Calle 54 doesn’t have the same language and class barriers that affect Buena Vista Social Club, and represents musicians in a different set of circumstances. Calle 54 highlights a set of musicians active in Latin/Afro-Cuban jazz — the flavors of carribbean-jazz fusion that evolved along with early jazz, flowered in new york in the 40s and 50s, in the 70s alongside the salsa craze, and continues until today. Most of the musicians in Calle 54 had continuous careers, with ups and downs. Gato Barbieri had retired for apparently mostly personal reasons, but had already returned to performing by the time the film was being made.

Prueba is from Spain; the film is in Spanish with subtitles. He interviews the musicians, who, in snippets in the movie, and especially in an excellent DVD add-on, talk about their careers and the history of the music they play. In separate interviews, the musicians share similar answers, about the African traditions and rhythms that are the foundation of the music, and the bidirectional networks of collaboration among US and Carribbean musicians that formed this fusion.

I love the stories of the interactions among the musicians and traditions – how the different African traditions contributed to Cuban music – how New Orleans musicians would come to Cuba, march in the parks during the day, and play in the clubs at night – how Dizzy Gillespie sought out Cuban musicians to forge the style, and much later gave a phone call to the young starstruck Jerry Gonzalez to fill in for a missing percussionist. Gonzalez later took the lessons to found his own Fort Apache Band. Wherever one looks, cultural collaborations are always (always) more inter-related and weirder than romantic myths.

One of the artistic tensions shown in the film is between the musicians desire to play popular dance music, and more musically challenging jazz. Big band leader Arturo O’Farrill says that he was motivated to write music by the desire to add compositional interest to the simpler structures of Cuban popular dance music; Paquito D’Rivera talks about the tension in Irakere between making hits like Bacalao con Pan and jazzier, more complex pieces. Another of the tensions is in the musicians efforts to combine traditions with integrity and interest; Prueba talks about attempts to combine flamenco and jazz with varying success.

I am no expert or connoisseur; and have no special technical or cultural background, I just listen, but to my ears the tensions result in plenty of interesting and lively music for further listening. My favorite moments were the piano duet between formerly estranged father and son, Bebo and Chucho Valdes; the Paquito d’Rivera band, the Fort Apache band. Gato Barbieri’s tone just kills but his band wasn’t that interesting to me. Eliane Elias’ piano doesn’t do much for me. I can’t tell if it’s her playing, my taste, or if my ears have been ruined by film scores and hotel lobbies.

Like Buena Vista, calle 54 is mostly apolitical on the surface. The part that seems to me like visible social commentary is the interviews of the Gonzalez brothers of the Fort Apache band, named after the rough Bronx neighborhood they grew up in. The myth in American media is that the neighborhood was an irredeemable wilderness; the reality was more complex, with lively culture, poverty, and social problems co-existing; in the DVD add-on, Andy Gonzalez tells the story of an attempted mugging at a subway station coming home from a gig, by some junkies who were after his bass, the junkies were chased off by the sight of a cop car. His brother tells a story about leading a pack of young teenagers sneaking into a local amusement park to hear a jazz band play.

in summary, Buena Vista Social Club is more of a fable, and Calle 54 is more of a music film. In part because of the film-making, and in part because of circumstance, in Buena Vista the story is largely about the musicians, and in Calle 54 the story is largely told by them. I liked and recommend watching them both, and then raiding the discographies if you haven’t already been big fans.

Trust is contextual

As healthcare reform passes, I’m having a strong but respectful disagreement with a friend over twitter on the nature and impact of this change. I know him through professional circles and I’d recommend him for jobs; and I’d trust him to pick me up on the highway if I was stuck in his town. But I wouldn’t trust him for advice in political matters. There are friends whose political judgement I trust and seek out, but whom I wouldn’t trust to take me to a concert or a movie I’d like. There are people who’s advice I’d take personally but not economically, and vice versa. Trust is faceted. Trust is contextual.

Craig Newmark, who combats untrustworthy behavior every day on Craigs list, believes that distributed trust is the next big problem for the internet to solve. If Google, Facebook, and Amazon got together, they might be able to address this problem of untrustworthiness online. Jay Rosen, who I trust and think is wise about many issues, agrees a distributed trust network is needed.

I am a huge fan of distributed solutions to many problems, but not this one. Trust is contextual. Even trust within a specific online service can’t be generalized. The other week I was scouring the music recommendations of a prolific Amazon reviewer with deep musical taste in areas I like. That same reviewer’s opinions about religion and politics are 180 degrees away from mine. He could buy me a recording sight unseen and I’d probably love it, but I couldn’t read most of his book recommendations without throwing them.

Bruce McVarish believes that the trusted circle will start with people we trust in real life. But even in a close personal circle, trust is contextual. Even among my family and closest friends, there are different people I would trust for different things.

Trust can be extended along specific and faceted lines. In the important area of transactional trustworthiness, ebay-style ratings are critical. A distributed solution for transactional trustworthiness could be quite useful. It would be handy to have distributed trust metrics that could be extended across services in a given domain – I’d love to follow the Amazon music recommender across his various music services on and Spotify and so on. But trust in any one domain doesn’t extend to other domains. Someone might be a meticulously reliable seller of used books and electronics, but they might be a horrible filter for news, which is the area that Jay Rosen cares about; or they might be personally unkind, so wouldn’t make personal trust list.

There is no general distributed trust solution to be had. Trust is always contextual.


Update 1: a few more thoughts, in response to some questions on Twitter and in comments

Trust, here, is the inverse of reputation.

For the facets of trust where there might be some tractable technology help, the facets need to be addressed with different kinds of technology augmentation.

* In the area of transactional trust – do I trust you to deliver this used book on time in the condition you promised – trust may be represented as a number. If your trust score is 99/100 I’ll buy the book. If your trust score is 62/100 I won’t buy the book from you.
* But in the context of opinions and tastes – would I like your movie recommendations – the number is not so useful by itself, but only as an indicator about similarity of opinion. So, if an algorithm says that our tastes are 75% similar, then I may want to subscribe to your movie recommendations.

So, for transactional trust, a distributed trust solution might aggregate a reputation score. For opinion trust, a distributed trust solution might calculate a score based on actions in multiple services, but then aggregate the actions (like movie recommendations) across services.’s music similarity scale works along these lines already, based on aggregating listening information. Last allows users to scrobble their listening across services, shows how similar your taste is to others, and then allows you to explore the listening of these others.

Also, to anticipate another question, in social media, recommendations and other such trust-building actions are social gestures, not just quantitative ones. I recommend a movie to show off my own taste, to be generous to friends, to amuse, to give a gift, to express my similarity or difference with a public, and so on. And within this social dynamic, it would be helpful to be able to identify people who have enough similarity to want to share these gestures; to aggregate the identification across services, and to aggregate the recs across services.

Update 2 in response to comments from Thomas Vander Wal and Charles Green.

It’s important to consider that the set of circumstances where technology help with trust problems is really constrained. Both Vander Wal and Green spend much of their time helping with people to communicate better in groups – in these situations, the responses are largely about people, customs, culture, values, leadership, facilitation, tummeling. Tools are small part of the response, compared to the human aspects. Charles Green summarizes well: “The things that can be scaled up through numbers on the internet are important, but limited.”

From the perspective of technologists, numbers and tools are hammers, and social concerns may look like nails. Technologists tend to overestimate the set of problems that can be addressed by technology. Part of the job of those of us working in social software and social media is to be analytical enough, and humble enough, to identify the things that are tractable with metrics and tools, and those things that need to be handled by people regardless of tools.

Even the frame of “problems to solve” that technologists bring is part of the problem, sometimes. Even with good will, people are always different, with different perspectives and interests. Buber, Levinas (and other purveyors of wisdom) remind us that people being different is a fundamental condition of life and an opportunity, not a problem. Sometimes there is no “solution” because problem is often the wrong frame.

Eve’s revenge – the snake made us human

In The Fruit, the Tree, and the Serpent: Why We See So Well, anthropologist Lynn Isbell makes the case that humans evolved distinctive capabilities to see and to communicate, in response to snakes, the most dangerous predator of our primate ancestors. The book marshalls evidence across a range of disciplines: neuroscience, primate behavior, paleogeography, molecular biology, and genetics to make the argument.

Isbell’s argument sounds like “Eve’s revenge” against the argument in Terrence Deacon’s Symbolic Species. While Isbell argues that “the snake did it”, the Deacon argues that “Adam did it” – humans’s understanding of symbols evolved to express the sexual ownership of women by men. In his book, Harvard neurologist and anthropologist argues that the understanding of symbols is the main differentiator of human intelligence. Humans invented an abstract symbol and group ritual – such as wedding rings and marriage cermonies – to mark the fact that a woman is the exclusive sexual property of a man. This allowed humans to live co-operatively in groups (which enables more efficient hunting and gathering).

Both stories are fascinating assemblages of scientific evidence making arguments about the origins of distinctively human capabilities. But, as I said in a blog post about the Symbolic Species, Deacon’s argument skips over alternative explanations of the same evidence. The ability to see beyond immediate evidence to consequences remote in time and space can be explained with Robin Dunbar’s hypothesis of gossip, or storytelling in general. Deacon himself argues that language is overdetermined; there are so many advantages that it’s hard to tell what came first.

I enjoyed Deacon’s book for the evidence it brought about how the human brain processes language, and I look forward to reading Isbell’s book, which I read about in the Atlantic review, for the research she assembles about the evolution of vision and cognitive capabilities. But I strongly suspect that I’ll think about Isbell’s argument what I thought about Deacon’s – there is evidence for it, but there are also many other ways to explain the path of evolution, and no solid way to prove these explanations of the distant past. We’re humans, so we search for causes and tell origin myths, even when we’re using the tools of science.


to be Dunbarred (definition): to have so many people as friends/followers in a social network that one can no longer easily pay attention to new people. British anthropologist Robin Dunbar proposed that there is a limit to the number of people with whom one can maintain social relationships, and that the limit is a function of relative neocortex size; humans can maintain larger relationship networks than other primates, for example. See Wikipedia Dunbar Number. There are various hypotheses about what the limits might be and what they mean – the core idea is that we’re limited in the number of people we can connect with.

Update: the Urban Dictionary had a definition of to dunbaras a transitive verb, posted by judiec on Feb 4, 2010, meaning to unfriend someone because you already have too many friends.

To intentionally remove someone from your circle of friends by avoiding contact, hiding or in extreme measures moving away. This will normally be necessary due to dull behaviour, fun sapping or heightened arrogance.


So I submitted the intransitive definition, to be overloaded. Thanks, Audrey for the Urban Dictioary reference.

Social media subcultures

Bora Zivovic recently pointed out an interesting study by Christina Pikas, which looked at link and comment connections between science blogs to look for clusters in the community of science bloggers. Her study revealed a few interesting things. Scientists read blogs by discipline, not just their own subdiscipline, but a variety of related subfields. In addition, there is a cluster of science blogs by women across multiple fields who comment on each others blogs especially often.

Ever since the Clay Shirky popularized the fact that there are power laws in the distribution of attention in social media, which focused attention at the top of the fame pyramid I’ve been wanting to see more studies like Pikas, which focus instead further out on the curve, closer to the long tail. Instead of showing who’s at the top of the pyramid – information that many people already know, use tools to discover and understand subcommunities, where a lot of the interesting human interchange, culture, and creativity. Within these subcultures, a closer look can show the distinctive attributes of the community itself. It’s helpful that Pikas herself is a member of the science blogging community – she knows enough about it to interpret the results of the social network analysis in a meaningful way.

A lot of the interesting dynamics in social media is at the level of subcommunity and microfame. More like this please.

Diversity and social design

In a post last week, Adrian Chan called out claims that competitive motivations are predominant in human nature.

One such quote comes from Louis Gray, writing on the need for more meaningful metrics than followers. “Humans have this innate sense of need to be ahead of all others, to measure themselves, and deliver some level of self-assigned worth thanks to what are questionably valuable statistics.”

Adrian rightly observes that, “it’s not humans or human nature that are the cause of this. It’s systems and the design of social experiences and systems…. viewed empirically, societies around the world are organized in wonderfully different ways, manifest in a tremendous range of culturally diverse traditions and pastimes.”

People in different cultures have different preferences with about the expression of value and status. For example, the design pattern wiki for O’Reilly’s Designing Social Interfaces describes differing cultural preferences for ratings. “The very notion of a binary, black-or-white, love-it or hate-it ratings system may not be a natural fit for some cultures. Many locales have stated a preference for ‘shades of gray’ in a polarized scale. (Note this is subtly different than Star-ratings, which imply “I like it exactly this much.”)?

To remedy the situation, Adrian calls for a better understanding of social dynamics. The current understanding gap is exacerbated by a lack of diversity in social design; consciousness of diversity among the user community, and actual diversity among designers. The lack of diversity contributed notoriously to the Google Buzz privacy debacle – Google tested Buzz internally for six months, but Google’s employees – mostly young male engineers – were not concerned that the system exposed users’ list of email contacts. The social networks of Google employees were more homogenous and harmonious than the social networks of the gmail user base. As soon as the product was released, users with different experiences – people who were harrassed by stalkers – consultants who had confidential client lists – quickly objected to this disclosure.

Embedded in the practice of software design is the notion that “you are not your user”; designers need to consider the way that users preferences may be different from their own. Demographic diversity is the most visible sort of diversity – differences in gender, age, ethnicity, geography. Diversity isn’t just a matter of demographics; there are differences in personality and temperament that lead to different social interaction preferences. As social software and social media mature, designers and implementers will benefit from a greater understanding of differences in among users.

The need for diversity in social design is another reason to advance the development and adoption of standards. A monoculture of a few dominant vendors is less likely to generate a diversity of social software affordances that that will meet the variety of needs among social software participants. The spread of standards will make it more feasible to have a diversity of suppliers, creating a diversity of services meeting different needs and preferences.

The revival of groups in the age of the network

In a recent blog post, David Weinberger writes about how networks have surpassed groups in recent years, as ways of defining social connections online. “In the past decade, we’ve gone from talking about social circles to social network. A circle draws lines around us. Networks draw lines among us.” Social network messaging, where communication centers around the individual user (such Facebook and Twitter), have rocketed to prominence, far ahead of group-based tools (such as found in Yahoo Groups and Google Groups, other age-old forums, and special-purpose tools such as MeetUp).

Weinberger implies that groups are obsolescent: “(Yet more evidence — as if we needed it — that networks are the new paradigm. Bye bye, Information Age!)” Networks are more visible and addressible now, but I don’t see groups becoming obsolete. As networks grow, groups are poised for a major comeback, as a way of expressing context within networks.

Scale and context

One of the reasons that social messaging networks have surpassed group forums is that networks scale around the individual. When you join a group, the level of noise depends mostly on other people – when the place gets too popular, the experience degrades for individuals. In a network, each person controls who they friend and follow, and this puts the limit under the control of the individual.

But networks eventually scale out too. The number of people to friend and follow is under your control, but subject to social pressure and information greed – like chocolate you can get too much of a good thing. Keep adding friends, followers, and eventually there is too much information and not enough context.

The solution to too much information is more context. As Clay Shirky says, there is no such thing as information overload, only filter failure. One of the most important ways of filtering is adding context. Context helps people focus on who and what they care about it, when they care about.


Lists are a way of putting followers and friends into context that is centered around the individual. With the list features in Twitter and Facebook, each person can organize others into sets, using their own personal taxonomy. Lists help individuals manage attention in personal context. Twitter lists have a tiny bit of sharing – it is possible for one person to subscribe to another’s list. A list that is very popular could conceivably provide a shared view of a set of people. But there is no collaborative ability to curate lists or nominate oneself for a list.

Because lists are personal, they don’t create shared identity or enable shared action, which are powerful drivers of context. This is where groups come back.

Groups and identity

As Stowe Boyd and Adrian Chan remind us, identity is socially constructed in social context. Now, the assembly of a social context doesn’t require a formally defined group. Social context is shaped by people’s interactions and mutually recognized signs of affiliation, not by defined membership. In an open network (Twitter), social ties can be inferred from patterns of tag use and replies, more strongly than than mere follow lists, and despite the fact that there is no official group. Networks of replies and posters to a common tag become familiar faces. For example, on Twitter I’ve recently stumbled upon an informal network of Icelandic musicians and their fans.

But people often want persistent affiliation, recognition, and communication in groups. This is deeply human; basic traits that anthropologists catalogued when they decided that the behaviors of people were interesting to study. Within networks there is a basic need for a groups to express a greater level of affiliation, recognition, communication, focus.

Groups and action

Groups are handy for affiliation and shared identity; they are necessary for sustained action. Networks can be very effective for ad hoc action. Think about the way that the call for donations to help with Haiti emergency response spread rapidly on Facebook and Twitter. But to coordinate action over time, you need ongoing communication and longer sequences of actions.

In open source software development, the classic model of self-organized coordinated action in the internet age, a new project sets up a code repository, mailing list/forum, a wiki, and an IRC channel for ad hoc synchronous communication. The basic toolset for coordination includes group collaboration. The best practices in internet self-organization allow for increasing levels of organization, starting at a very lightweight level, where participants can read information, start to ask questions, and make small contributions, on to very high levels of dedicated contributions.

By enabling groups to form within larger networks, people get the benefits of a larger network, with more manageable, lightweight communication, while also being able to communicate and collaborate more deeply with a set of people with shared interests and goals.

Focus, not privacy
Often people consider the topic of social sharing in terms of “privacy”. The information overload symptom of “oversharing” is seen as a privacy problem. As Stowe Boyd and others observe, the issue of oversharing not primarily about information should be kept hidden, and much more about who to share with in what context. Even if you don’t care who knows who else knows your workout routine, fellow fans of rowing or weight-lifting might care more than other friends and colleagues.

Groups frequently aren’t private – in fact, they are more useful for many purposes if potential participants can easily find them, look around to see what’s going on, and join if they are interested. FriendFeed groups were quite popular among scientists and journalists online. Most of these groups were publically listed. Users could choose to join them. Another convenient setup is groups where a member can request to join, and a moderator needs to approve applications.

In most cases, the goal of a group isn’t to keep information secret – it’s to allow people to affiliate, to collaborate. And to focus their attention and communication within these defined social contexts.

Groups and networks – summary

In summary, I don’t think it’s true that the rise of networks is going to wash away groups. Groups and networks are complementary. Networks help people get to know other individuals, and to manage attention by constraining the number of people to follow. Groups help people focus attention, share identity, and collaborate more deeply within networks.

As ReadWriteWeb describes, a big part of the solution to information overload is increased context. And groups are key to re-establishing context in the network era.

Learnings about web ratings systems

“The Wisdom of Crowds” is one of the driving principles of Web 2.0. The idea, explored in James Surowiecki’s influential book, is that decisions made by large numbers of people together are better than decisions that would have been made by any one person or a small group. This principle has powered the wide adoption and success of tools including including Google, collaborative filtering, wikis, and blogs.

One common technique, following the Wisdom of Crowds principle, is the use of ratings. The hope and expectation is that by enabling large numbers of people to express their opinion, the best will rise to the top. In recent years, rating techniques have been put into practice in many situations. The learnings from real-life experience have sometimes been counterintuitive and surprising.

The failure of five-star ratings

Many sites including Amazon, Netflix, and Yahoo! used five-star ratings to rate content, and this pattern became very common. Sites hoped that these ratings would provide rich information about the relative quality of content. Unfortunately, sites discovered that results from the 5-point scale weren’t meaningful. Across a wide range of applications, the majority of people people rated objects a “5” – the average rating across many type of sites is 4.5 and higher. Results from YouTube and data from many Yahoo sites show this distribution pattern.

Why don’t star ratings provide the nuanced content quality evaluation that sites hoped for? It turns out that people take the effort to rate primarily things they like. And because rating actions are socially visible, people use ratings to show off what they like.

How to use scaled ratings effectively

So, is it possible to use scaled ratings effectively? Yes, but there needs to be careful design to make sure that the scale is meaningful, that people are evaluating against clear criteria, and that people have incentive to do fine-grained evaluation. Examples of rating scales with more and less clear criteria can can be found in this Boxes and Arrows article – the image from that article is an example of a detailed scale.

There are tradeoffs between complexity of the rating criteria and people’s willingness to fill out the ratings. Another technique to improve the value of scaled ratings is to weight the ratings by frequency and depth of contribution, as in this analysis by Christopher Allen’s game company. This techniques may be useful when there is a relatively large audience whose ratings differ in quality.


The simpler “thumbs up” or “like” model, found in Facebook and FriendFeed has taken precedence over star ratings systems. This simpler action can surface quality content, while avoiding the illusory precision of five-star ratings. The vote to promote pattern can be used to surface popular content. This technique can be used in two ways – to highlight popular news (as in Digg) or to surface notable items in a larger repository.

Several considerations regarding the “like” action: this sort of rating requires a large enough audience and frequent enough ratings to generate useful results. In smaller communities the information may not be meaningful. Also, the “like” action indicates popularity but not necessarily quality. As seen on Digg and similar sites, the “like” action can highlight the interests of an active minority of nonrepresentative users. Or the pattern can be subject to gaming.

Another concern is the mixing of “like” and “bookmark” actions. Twitter has a “favorite” feature that is also the only way for users to bookmark content. So some number of Twitter “favorites” represent the user temporarily saving the content, perhaps because they disagree with it rather than because they like it! Systems that have a “like” feature should clearly differentiate the feature from a “bookmark” or “watch” action.

The risks of people ratings

Another technique that sites sometimes use, in the interest of improving quality and reliability, is the rating of people. Transaction sites such as Ebay use “karma” reputation systems to assess seller and buyer reliability, and large sites often use some sort of karma system to incent good behavior and improve signal to noise ratio.

The Building Reputation Systems blog has a superb article explaining how Karma is complicated. The simplest versions don’t work at all. “Typical implementations only require a user to click once to rate another user and are therefore prone to abuse.” More subtle designs still have an impact on participant motivations that may or may not be what site organizers expect. “Public karma often encourages competitive behavior in users, which may not be compatible with their motivations. This is most easily seen with leaderboards, but can happen any time karma scores are prominently displayed.” For example, here is one example of karma gaming that affected even in a subtle and well-designed system.

Participant motivations, reactions, and interactions

When providing ratings capabilities for a community, it is important to consider the motivations of the people in that community. In the Building Reputation blog Randy Farmer talks about various types of egocentricand altruistic motivations. Points systems are often well-designed to support egocentric motivations. But they may not be effective for people who are motivated to share.

Adrian Chan draws distinctions between the types of explicit incentives used in computer games, and the more subtle interests found in other sorts of social experiences, online and off. People have shared interests; people are interested in other people. The motivations come not just from the system in which people are taking these actions, but from outside the system – how people feel about each other, how they interact with each other.

In a business environment, people want to show off their expertise and don’t want to look stupid in front of their peers and superiors. They may want to maintain a harmonious work environment. Or in a competitive environment, they may want to show up their peers. These motivations affect the ways that people use ratings features as well as how they seek and provide more subtle forms of approval, like responses to questions in a microblogging system.

Thomas Vander Wal talks about the importance of social comfort in people’s willingness to participate in social systems, particularly in the enterprise.
People need to feel comfortable with the tools, with each other, and with the subject matter. The most risky form of ratings, direct rating of people, typically reduces the level of comfort.

Depending on the culture of the organization and the way content rating is used, content rating may feel to participants like encouragement to improve quality, like a disincentive to participation, or like an incentive to social behavior that decreases teamwork. Even with good intentions and thoughtful design, the results may not be as anticipated. In that case, it is important to monitor and iterate.

Scale effects

The familiar examples of ratings come from consumer services like Amazon, Netflix, and Facebook, with many millions of users. With audiences as large as Amazon’s, there are multiple people willing to rate fairly obscure content. In smaller communities, such as special interest sites and corporate environments, there are many fewer people: hundreds, thousands, tens of thousands. While the typical rate of participation is much higher – 10-50%, rather than 1-10%, that is still many fewer people. With a smaller population, will there be enough rating activity to be meaningful. If an item has one or two ratings, what does this mean? Smaller communities need to assess whether the level of activity generates useful information.


Ratings and reputation systems can be very useful at surfacing the hidden knowledge of the crowd. But their use is not as simple as deploying a feature. In order to gain value, it is important to take into account lessons learned:
* Think carefully about the goal of the ratings system. Use features and encourage practices to achieve that goal
* Use an appropriate scale that addresses the goal
* Consider the size of the community and the likelihood of useful results
* Consider the motivations and comfort level of the community and how the system may affect those motivations and reactions

Then, evaluate the results. The use of a rating system should be seen not like a “set and forget” rollout, but as an experiment with goals. Goals may include quantitative measures like the volume of ratings and the effect on overall level of contribution, as well as qualitative measures such as the effectiveness of ratings at highlighting quality content, the effect on people’s perception of the environment, and the effect on the level and feeling of teamwork in an organizational setting. Be prepared to make changes if your initial experiment teaches you things you didn’t expect.

For more information

The Building Reputation blog, by Randall Farmer and Bryce Glass, is an excellent source of in-depth information on this topic. The blog is a companion to the O’ReillyBuilding Web Reputation Systems.

Other good sources on this and other social design topics include:
* Designing Social Interfaces book and companion wiki, by Christian Crumlish and Erin Malone.
* Chris Allen’s blog
* Adrian Chan’s blog