book review: Delete

a very thoughtful book about the role that remembering and forgetting have (supposedly) played throughout human history, and why we are for the first time at a point where remembering, not forgetting, has become the default.  the author finds this problematic for two main reasons.

the first has to do with power.  despite what is said, information is not power — it’s the ability to control and restrict access to information that is power.  Mayer-Schonberger worries about the extent to which we may, whether we know it or not, lose control over pertinent information about ourselves, including more troublesome features like our personal information, as well as our tastes and opinions that outside agents use to great extensive — and unending — dossiers about us.  we are not in a great position to control who has that information and what they do with it.

the second has to do with time.  Mayer-Schonberger sees forgetting — as frustrating as it may be — as a blessing in disguise in that it lets us freely reconstruct our memories — which are far from perfect — as we progress through our lives, choosing to focus on (and thus retain) certain memories and not others.  what happens when we have perfect (not now, but someday soon) recall, through digital technology, of all of our past decisions and events?  will we, like the few individuals who do have perfect memories, be frozen in indecision, constantly replaying the past and inept at acting in the present?

the author then takes a look at a number of possible solutions, from digital rights management to information abstinence and other suggestions, but decides that these are not “silver bullets” for confronting the problem that we face today.  however, he does not reject these, and in fact he hopes to see many of them in place.  his concerns are pragmatic and focused on where we are now, and he does not denigrate other suggestions (many from his peers) in self-promotion.

his ultimate solution is to use meta-data to place expiration dates on information that we store or share with ourselves, in addition to supporting programs in our technology (e.g., storage devices) and online tools (e.g., browsers) that automatically erase information about our searches after a time period — a period we may have some control over.  we could thus choose what types of queries we want google or amazon to remember indefinitely, and which ones to forget after some time.

overall, Mayer-Schonberer’s suggestions are very interesting ones, especially since they are rooted in the technology as we have it today, although there would need to be some pressure on companies to implement some of these changes to help people make the transition (he takes humans as they are, with imperfect knowledge and motivations, and that’s another strength of the book).  it’s  not entirely certain how all of these suggestions would play out — and how inconvenient it would all be.

in the case of putting a shelf-date on personal photos shared online — which is one of the cases he uses to illustrate the extent to which information online can come back to haunt us — it seems difficult to know how you would decide what expiration date to put on something like that.  plus, if you consciously put a date on a photo that could get you into some trouble — maybe you should think twice about sharing that photo.  indeed, in the case he uses — someone being denied a teaching certificate because of a questionable photo on myspace — a photo like that could be found at any point before the expiration date and the individual could still face unforeseen consequences — after all, that’s the point of the story: it was a seemingly harmless photo that the individual never thought could be used against her.

but again, the strength of the author’s position — and of the book overall — is that he is not out to establish THE plan for confronting these concerns; rather, he works within the existing frameworks to offer a novel solution that could help fill in some of the gaps and — in keeping with the concerns of the book — return to individuals some of that control over information and help shift the balance back to a society in which forgetting — and not remembering — is once again the default (even if just barely).

book review: From Gutenberg to the Global Information Infrastructure

this was a very disappointing book.  although it does contain some insightful discussions, so much of the book is devoted to discussing the potential definitions of the relevant terms (“digital,” “library,” “access”) while not really offering anything novel — she ends up going with an essentially intuitive understanding, and so, so much time has been wasted.

what disappointed me most was the promise of a real analysis of information infrastructure and “the social life of information” based on historical precedents.  after all, the title does suggest such a scope.  but this is lacking, apart from a few casual, generalized references.

moreover, what this book lacks in vantage is not made up for in vision.  it ends up being one of those books where the author discusses many of the problems we’re facing and offers, as a solution, a general discussion of what other people are doing (but no specs) or what other people imagine might happen, tying it all together with generalized thoughts and “we should think more about this” remarks.  don’t write a book about what you think we should do or what might be a solution!  do the work and report back if it seems to be working!  don’t use a book to talk about how we should talk about a problem!

while I don’t think she’s ever blatantly wrong, she is irredeemably boring, which, in a book with few other saving graces, is like being kicked when you’re down.

filter & focus

over at the Edge’s big question, David Dalrymple muses, “Filtering, not remembering, is the most important skill for those who use the Internet.” as I continue to think about how and what we should learn in an increasingly information-overloaded world, I tend to think that what we should be learning is how to think critically, how to access information, and how to sift out the rubbish. that’s not to say that we shouldn’t still cover the basics of a traditional education (language & literature, science, math, history, etc.), but it would significantly alter our approach to how we teach and what we expect students to know or to be able to do. I think memorization — while still relevant and unavoidable in our daily lives, and especially in various careers — is a virtue of bygone days. what we need to know to live our lives and to function effectively we will invariably learn (or suffer the consequences), but it seems that people are likely to drown when flooded with so much information and they have no idea how to stay afloat.

perhaps more important than trying to learn a basic science (which will soon be forgotten) would be, in addition to a basic overview (for which students would be responsible) would be learning the bigger picture and the methods by which scientific conclusions and consensuses are reached. more important than a few courses on Western history (and maybe only one devoted to the rest of the world), during which students memorize dates and skeleton outlines would be learning the principles of historiography and how to go about evaluating claims and doing historical research (again, along with a broad picture of world history and those essential “plotting points” essential for building more in-depth knowledge).

I don’t envision a future in which no one will learn anything anymore, relying on our stored information (that we can access whenever we want) — and even if we do more than now, so what? I possess very little basic survival information, but I think it would be pointless to still teach our children all the necessities, starting with Stone Age tool making and hunting, all the way to the present. we’re fortunate that information is so readily available — but what we need are the tools to learn how to use it effectively (and the continued incentive to go beyond the surface and analyze what we actively research or passively receive).

anyway, the author also weighs in on the extent to which our ability to focus may be a better indicator of job performance than what we have heretofore learned.

Before the Internet, most professional occupations required a large body of knowledge, accumulated over years or even decades of experience. But now, anyone with good critical thinking skills and the ability to focus on the important information can retrieve it on demand from the Internet, rather than her own memory. On the other hand, those with wandering minds, who might once have been able to focus by isolating themselves with their work, now often cannot work without the Internet, which simultaneously furnishes a panoply of unrelated information — whether about their friends’ doings, celebrity news, limericks, or millions of other sources of distraction. The bottom line is that how well an employee can focus might now be more important than how knowledgeable he is. Knowledge was once an internal property of a person, and focus on the task at hand could be imposed externally, but with the Internet, knowledge can be supplied externally, but focus must be forced internally.

book review: Infotopia

this book, as its subtitle indicates, is about the production of knowledge by many minds.  but the book is less about the fact that many minds produce knowledge than about the ways in which information that is dispersed among many minds can be accessed and the conditions under which those varying methods work best.  under discussion are surveys/polls, deliberation, markets, wikis, open source software, and blogs.

so, for instance, he starts off the book talking about the surprising ways in which large groups of people can outperform individuals when answers are averaged out.  often the average answer — when guessing the weight of some object, when trying to correlate body weight with gender — is not only better than the best individual answer, but also better than what a supposed expert can offer.  to be sure, aggregating information like this only works under specific conditions, say, when it is reasonable to presume that people might have a general idea about something.  it would be useless to rely on the statistical responses of people for information not privy to most people, say, the year of some lesser known historical event or the name of someone’s pet (unless that someone is famous, maybe).

the reason that this works, Sunstein explains, is due to the Condorcet Jury Theorem, which states that the probability of arriving at a correct answer increases as the size of the group increases provided that there is greater than a 50% chance that people will arrive at a correct answer.  the more people you have, the closer you approach to 100%.  this is the reason why “ask the audience” usually works well in Who Wants to Be a Millionaire? — because there’s a decent chance that some people know the answer, or at least can correctly rule out false answers.  in these types of situations, it is beneficial to rely on the responses of a large group of people (as large as possible, in fact) to increase the chances of arriving at the correct answer.

the flip side of this math, however, is that if people have less than a 50% chance of knowing the correct answer — again, when asking about information not widely disseminated — the probability of arriving at the correct responses approaches 0% as the group increases.  so clearly this isn’t always (or even often) the best way of arriving at the truth.

the second method under review is deliberation.  Sunsstein is open about giving deliberation a bad rap here not because it is entirely inefficient, but because it is so often assumed to be the ideal way of accessing dispersed information and thus the truth.  deliberation lies at the heart of many practices in this country, from trials by jury to our deliberative democracy.  the problem with deliberation, in short, is that it doesn’t work very well.  Sunstein offers a number of reasons for this, owing to some of the natural shortcomings of the human mind (some familiar terrain after reading Kluge) and to particular behavioral phenomena in group settings, such as the general “groupthink” idea, along with informational cascades (when people factor into their responses the likelihood that other people, who may hold a different opinion, would be wrong and so answer or vote not purely on the basis of information but on what everyone else appears to know as well) and the many pressures on individuals to preserve group harmony (or their own status) by not offering information they may have that goes against the conventional group wisdom.  in experiments, people also tend to accord more authority to people in higher positions (including class, gender, and race — even if those social statuses are irrelevant to the immediate context) and to ignore others, regardless of the value of the information.

in one particularly illuminating example, the author discusses an experiment in which individuals of a group are asked to vote for candidates in an imaginary election.  the experiment is set up in such a way that Candidate A is clearly the most fit choice for the position.  when group members are all given about 2/3 of the relevant information for the candidates, the deliberation usually results in the correct choice of Candidate A (a statistical improvement over the initial poll of individuals — so here, deliberation helped).  however, when the members are all given 2/3 of the information about the other candidates, and the information about Candidate A is dispersed among individual members (even if the total information is more than in the previous scenario), the groups fail to access the relevant information contained by some of its members.  as a result, they end up choosing one of the demonstrably inferior candidates.  moreover, the percentage of votes for Candidate A fell after deliberation.  why?  because the information favoring the wrong candidate is that which is held by all the members — a phenomenon aptly called “the common knowledge effect.”

the major concern here is that deliberation groups often fail to access the relevant information held by some of its members because of the tendency to favor (and focus on) information shared by all rather than on individual perspectives, even when there was no evident (or stronger than usual) “status” issues or instances of social pressure on conforming to group opinion (indeed, there was no group opinion until the hypothetical information was given out).  in other experiments, the success of deliberation groups was also dependent on whether the group members were “primed” to think that arriving at the correct answer was important, as opposed to priming them for getting along.  this is cold comfort when thinking of juries and governmental deliberation.

this is not to say, however, that deliberation never works — obviously it worked in the first part of the experiment.  indeed, deliberation groups can perform as well as their best member, and sometimes they can even outperform their best member when pieces of relevant information are dispersed and the information, together, helps the group arrive at the correct answer.  but deliberation is best limited to instances when an answer is readily available (like problem solving) or “eureka” problems — when the correct answer can be identified by all as soon as it is made apparent.  on more ambiguous matter — say on social or moral issues, or anything involving ideology of whatever sort — deliberation groups are fairly terrible, often resulting in the amplification of previous biases (a well-documented event, familiar to anyone who’s ever been in a chat room or on a message board — or even among a group of like-minded friends, really).

Sunstein then moves on to markets — prediction markets, more specifically.  on the general level, the author discusses why online review sites (of movies, restaurants, products, etc.) have worked so well on the principle of a market and the establishing of a “price” of a particular commodity.  but what is most interesting is his discussion of more recent developments of prediction markets in which people place value (and trade stock) on the likelihood of a certain outcome — say, the winners of Oscars or the results of a political election.  surprisingly, these “markets” have often (but not always) outperformed even the best experts in their predictions.  the reasons why these markets work is that they provide an incentive for people with good information to put their money where their mouth is, resulting in predictions made by people who, in theory at least, have relevant information.  if you are concerned, as the author is, with how we most efficiently go about accessing widely dispersed information in society, then markets are often an excellent way of bypassing some of the social pressures and dynamics of deliberation groups.  these don’t always have to be (indeed, they often aren’t) open to the public and so can limit the predictions and trading to the relevant individuals.  so far, these types of markets have proved excellent within individual companies (e.g., Google and HP) at predicting what products will be the most successful or when a new product or program will be ready for distribution.  this new approach undermines conventional wisdom of a board of big wigs — who couldn’t possibly have access to all of the relevant information possessed by all the employees — making the decision from the top down.

to keep the rest of this brief(er), Sunstein then moves on to the various Web 2.0 developments in social media and information aggregation — including wikis, open source software, and blogs — and discusses their relative merits, as well as causes for concern.  as it turns out, unmediated forums for the sharing and refining of information have proved more effective than many feared.  that is not to say there are not problems with, say, wikis — indeed, Wikipedia is far better on some topics than others, and even then usually as a general guide, not the end-all authority — or blogs — here we can find some pretty terrible groupthink behavior, along with more than generous helpings of bullshit — but overall, they are very effective in ensuring that dispersed information sees the figurative light of day.  in fact, Sunstein discusses a few instances where information shared online by bloggers helped to correct statements made by political candidates (leading to apologies) or to debunk a phony document (leading Dan Rather to apologize and retire).

the book ends with a few discussions about the situations in which the various methods work best and a few suggestions about how groups and organizations can best make use of them.

overall, this is a very interesting book and fascinating information.  unfortunately, for even such a short book (225 pages), it was more repetitive than necessary and could have benefited from more individual case studies.  also, while I am tempted to say that this book is to groups what Gary Marcus’ Kluge is for the individual mind, this book is not nearly as entertaining and engaging as Marcus’, which is unfortunate because it certainly had the potential to be as captivating and perhaps even more relevant.

scientific truth in the age of Wikipedia

in February of last year, Seed magazine ran a brief article addressing the question of (concern about) scientific truth in the age of Wikipedia — that is, whether “the radical egalitarianism of the wiki undermine notions of scientific authority and consensus.”  obviously, as is pointed out, group consensus is not always the best way to arrive at objective truth.  some are concerned (indeed, outright frantic) over the extent to which this bottoms-up approach will undermine those hard-won facts of, say, scientific or historical research.

but the author of this article (T.J. Kelleher) argues that these fears are mostly unfounded, as “[i]t is not scientific rigor that is accommodating the wiki, but the wiki that is accommodating science.”  just think of areas of knowledge where no control test is possible (e.g., in historical research, where we are limited to the accident of history and the chances of discovery) or practical (e.g., large-scale scientific hypotheses about populations or, as the author proffers, the human impact on global warming).  here, consensus is often based on the best available knowledge and the most rigorous forms of analysis; often it comes down to what hypothesis best accounts for all of the data — and for this, consensus is key.

further, Clay Shirky — author of a book I’d like to read — argues that notions of authority and expertise are only “social facts,” and that works (and their writers) are considered authoritative only because everybody agrees that this is so.  and to be honest, when venturing into an area of knowledge with which I am not familiar, I will often reach out to recognizable landmarks and sources that I deem — from experience elsewhere and ongoing critique — to be authoritative (e.g., a particular author, a publisher, a university or institution affiliation, etc.).

when it comes to the scientific process, some believe that a source like Wikipedia will actually prove useful in its ability “to act as a chronicler and creator of repositories for science, to create a home for what we haven’t proved to be false and for what we think to be true.”  in other words, wikis can be beneficial to the notion of falsifiability — a key step in the scientific process.  thus bad ideas can easily set aside, and good ones can be continually refined, since the business of science, as some would argue, is more often than not establishing what is false, not necessarily what is true in some unchangeable sense.  to this end, as an MIT professor puts it, “the wiki model is ‘reasonably good’ at discerning, if not what is true, then what is false.”

one problem that remains, however, is the anonymity of the contribution and editing process.  not only is this a problem for managing the content of the site (are the contributors even qualified?), but it may also undermine the processes whereby we uphold notions of authority, “social fact” though it may be.  when someone stumbles across differing opinions or is left with an ambiguous answer to a query, how does that person go about seeking a more authoritative take on the problem?  if we cannot all become experts in the various fields of knowledge in which we might dabble — or into which we must wade (e.g., when seeking emergency medical information) — how do we learn where to invest that “capital” when it matters if knowledge production or maintenance is done anonymously?  you might feel slightly better about taking someone’s advice if they have widely acknowledge credentials and if their professional career depends on them being right.

some scientific organizations have sought to counter this problem by creating rival “wikis” that remove the veil of anonymity and involve an application and approval process that restricts an individual’s contribution to their certified field of knowledge.  and even here, the process of peer-review and consensus continues to play an important role.

personally, I don’t think that Wikipedia should be much of a threat to the scientific process or to fields of knowledge, provided that there is an understanding of where Wikipedia’s value lies.  if individuals turn to wikis for answers to every problem or as the end-all source of knowledge, unaware of how to go about verifying that knowledge and where to go for more in-depth understandings, I don’t blame Wikipedia — I’d blame any academic system that has failed to demonstrate how research is conducted.  many people in this country or woefully ignorant of the basic scientific process, not to mention principles of, say, historiography, but I don’t think that this is a new problem resulting from the development of more egalitarian forms of knowledge maintenance such as wikis.

also, I think that Wikipedia is a fairly reliable source — introductory source, that is — for many of the bodies of knowledge I’ve stumbled across.  and I am pretty sure that the rules have changed so that not just anyone, at any time, can make whatever change he or she would like with little to not policing.  so far I think that there has been substantial — and satisfactory — sifting to ensure fairly reliable information — often superior to (and far more practical than) traditional storehouses of knowledge such as encyclopedias.