Foraging on Quora: Making the site better for sharks, microplankton, penguins, sea turtles, bumblebees, albatrosses, spider monkeys, butterflies, honey bees, amoebae, and fruit flies

July 28, 2011 § 13 Comments

By: Edwin Kite

Disclaimer: I’m just a grad student in an unrelated field.

This is about information consumption on Quora. I won’t say much about the site’s social side.

People forage for information on Quora. It’s different from StackOverflow, where people dip in, and Wikipedia, with its sterile content. Foraging on Quora is more rewarding, because the quality of the thinking is occasionally very high. Foraging is structured by Quora’s feed – adjacent entries in the feed can lead down very different roads. The feed is tightly personalized, with a little randomness thrown in. Users who follow the feed’s cues will diffuse around Quora. Occasionally they will stumble into a new area. There is a natural length scale to their movements: the range of the people and topics they follow, because that’s what sets the breadth of their feed. This is like Brownian diffusion.

As an alternative, consider the albatross.

Albatrosses are long range ocean foragers. They are big – the biggest living thing that flies. So big you can strap a GPS receiver and a satellite transponder to one, and it barely notices. The data show [1] that the albatrosses do take many little hops, like Brownian motion. But occasionally they will take a bigger hop. And rarely, they will fly to the other side of the world in a straight line. In fact:-

where l is the length of the foraging flight, and P(l) is the probability that the hop will be that long.

This scale-free pattern (including the exponent -2) has since been found in sharks, microplankton, penguins, sea turtles, bumblebees, spider monkeys, butterflies, honey bees, fruit flies, and amoebae [2,3]. This is called a Lévy flight.

Lévy flights with exponent -2 are optimal for locating resources [4] when those resources are –

  • Randomly distributed
  • Sparse
  • Once visited, are not depleted, but remain targets for future searches (like a person, or a topic)

Sounds like Quora. By manipulating the environment, exponents between -3 and  -1 can be produced in ecological simulations. However, including a “crowding” effect (evaporative cooling?) does not affect the -2 exponent, at least not in mussels [5].

For a foraging Quora user, distant topics and users on the graph are also

  • Diverse
  • Rapidly changing
  • Altered by the presence of other foragers

Much of the interesting content is coming from people who pop up, write a few dozen excellent answers, and then either leave or start producing mediocre content. This makes the system dynamic. To find these nuggets in your feed, you need to rove far and fast.

It’s unreasonable for users to apply Lévy flight behaviour themselves.  This is because Quora is a murky fog of information, not a savanna. The “sightlines” are short, intentionally so, so users have to be tenacious and read many two-line answer previews before they find a truly interesting answer. To bring Lévy flight behaviour to Quora, the company would have to take the lead, by changing the feed.

Brownian motion makes sense for “campfire” social networks, and for Wikipedia. The kind of people who could make Quora great are allergic to sameness and want intellectual challenge. They need Lévy flights.

What would a Lévy-flight personalization system look like?

  1. Detect novelty-starved behaviour.
  2. Add a Lévy-flight component to that user’s feed. This is not just randomness, because increasing distance from the user’s preferences will always decrease P(l).
  3. Keep tabs on each user’s revealed preferences. Some will treat Quora mainly as a social network – keep “campfire” personalization for them. Some are foragers. Some will be in between. I guess the comments/PM ratio could be a tracer of this.
  4. Openness to experience is one of the Big Five personality traits, so the power-law exponent could be adjusted based on overall assessed openness.

Finally, all Lévy flights are super-diffusive – the mean-squared distance from the origin increases nonlinearly with time. In contrast, Brownian motion gives linear growth in mean-squared distance. Quora’s collection of questions and answers is growing faster than the diffusion of high-quality users away from their origin points. The resulting isolation is demoralizing. Celebrities, for example, are encircled by baying fans and rarely meet another celebrity. Lévy flights can fight this.

TL;DR: Quora’s feed favors campfire behaviour over foraging, which is bad for some users. Adding a scale-free factor would help Quora retain novelty-seeking users.

References

[1] Viswanathn et al., “Lévy flight search patterns of wandering albatrosses,”  Nature, 1996. (The Lévy-flight interpretation has been disputed, e.g. by Edwards et al., Nature, 2007. Everyone agrees, however, that most animal foraging has scale-free aspects and is not just Brownian, which is enough here.)

[2] Reynolds & Rhodes, “The Lévy flight paradigm: random search patterns and mechanisms,” Ecology, 2009

[3] Sims et al., “Scaling laws of marine predator search behaviour,” Nature, 2008.

[4] Viswanathan et al., “Optimizing the success of random searches,” Nature, 1999

[5] de Jager et al., “Lévy Walks Evolve Through Interaction Between Movement and Environmental Complexity,” Science, 2011.

(PM me if you can’t get hold of journal articles).

Shame on Business Insider, Gizmodo, et al. for quoting anonymous Quora answer as fact for pageviews!

July 8, 2011 § 19 Comments

By: Michael Sinanian

This blog has discussed Quora’s transformative effect on journalism before, bringing to light its disruptive potential on content commoditization and branding.

But over the fourth of July, a lot of this “news innovation” was effectively tossed out the window. The transgression? Prominent tech blog Silicon Alley Insider ran a story that was entirely derived from an anonymous Quora answer with no attribution what so ever.

Addressing a question about potential uses for Apple’s astronomically large cash hoard, the user wrote a lengthy answer that received over 570 up-votes and a long stream of congratulatory comments. [1]

Without reliable attribution, this answer should not be up-voted so enthusiastically and it certainly shouldn’t be repackaged and resold by the blogosphere as “news analysis.” This should be a lesson for both journalists and Quora users alike: well-worded answers that tell us what we want to hear aren’t necessarily good answers.

Other “respectable” sites also ran this story, including Forbes, Fortune, and Gizmodo, which went as far as to title their story “Is Apple So Far Ahead Because They Use Tech From the Future?”

Such titles are worthy of the term “link-bait” for their pure intention of securing re-tweets and re-posts to drive up page views, regardless of editorial value.

Although a win for Quora, whose content is now apparently so valuable as to be quoted without a shred of doubt by major tech news outlets, this is a sore disappointment for Journalism 2.0.

Silicon Alley Insider and all the other blogs that followed suit in reposting this should be ashamed of themselves. It’s stunts like these that make “blogging” a derogatory term in the face of “real” journalism.

Furthermore, as proud members of the Quora community, we must ask ourselves what’s permissible to up-vote. Sourceless answers like the one discussed may still be correct despite a lack of attribution, but that should be taken into consideration when deciding how to vote or sourcing it for a story.

This approval dilemma has arisen in other ways, too. We’ve all witnessed a highly up-voted answer simply because the content is well-worded or emotionally engaging, despite questionable correctness.

Unfortunately, Quora’s current approval mechanisms don’t have much granularity. We are only provided the blunt instruments of an up-vote, a down-vote, or a meaningless “thank you” note. It’s not possible to simultaneously up-vote for writing style but down-vote for attribution.

Despite such shortcomings, Quora does have a system in place that allows reviewers to specify why egregious answers need improvement in certain areas. Without opening this up to most of the user-base, however, content is still liable to be improperly credited on a wide scale.

For those of us who are enthusiastic users or people who use Quora for journalism, we must be careful. Our actions are these missing granular controls, and we must be cautious in their use.

[1] – The only attribution comes in the form of a few commenters who link to articles from earlier in the year that vaguely go over what the answerer talks about. This does provide some credence to the matter, but not enough for wholesale replication by other news outlets.

[2] – This is cross-posted from a personal Quora post which can be found here: http://www.quora.com/Michael-Sinanian/Shame-on-the-Blogosphere-and-Quora-Users

Where Am I?

You are currently viewing the archives for July, 2011 at The Quora Review.