Discussion 2.0: Personalization

[The fol­low­ing post is my sub­mis­sion to the Knight-Mozilla “Beyond Com­ment Threads” chal­lenge.]

The fol­low­ing are the core prob­lems with cur­rent dis­cus­sion systems:

  1. Trolls, acri­mo­nious peo­ple, and low qual­ity com­men­tary can drown out thought­ful dis­cus­sion and destroy a good community.
  2. Bias towards senior­ity: Deep insight is penal­ized if it comes from a new, unknown, or anony­mous voice. For exam­ple, on Quora, answer­ing a ques­tion one month faster than some­one else can lead to a rich-get-richer phe­nom­e­non where the old answer gets more upvotes because it is always shown as the top answer, and hence has more vis­i­bil­ity merely because it is older. Nepo­tism cre­ates arti­fi­cial fric­tion and a barrier-entry, because it is an effec­tive tech­nique for enforc­ing com­mu­nity stan­dards. But it has the down­side that it discriminates—gently or extremely—-against insight­ful new com­men­ta­tors and anonymity.
  3. Vot­ing sys­tems can be gamed by vot­ing rings and bal­lot stuffing.
  4. Vot­ing sys­tems can lead to “mob rule”.

How do we address all these core prob­lems in tra­di­tional com­ment­ing sys­tems?
How can we cre­ate an engag­ing sys­tem that most effec­tively pro­motes dis­cus­sion? Can we avoid nepo­tism and bias against new, unknown, and anony­mous com­men­ta­tors? How can we defend against basic trolling and vot­ing rings?
A next gen­er­a­tion dis­cus­sion sys­tem must address these core problems.

The core value of a dis­cus­sion sys­tem is to encour­age stim­u­lat­ing and engag­ing dis­cus­sion. We want a sys­tem that is fric­tion­less to par­tic­i­pate in: You can lurk for years and then jump in when you have some­thing great and insight­ful to say, and your voice is heard loud and clear. This is true democ­ra­ti­za­tion of discussion.

The solu­tion is per­son­al­iza­tion. A next gen­er­a­tion dis­cus­sion sys­tem is per­son­al­ized. Per­son­al­iza­tion makes dis­cus­sion more stim­u­lat­ing and engag­ing. Each user that reads and par­tic­i­pates gets a com­ment thread that is sorted by per­sonal rel­e­vancy. Irrel­e­vant com­ments are hid­den by default, but can option­ally be viewed. Per­son­al­iza­tion is tuned to pro­mote dis­cus­sion that the user finds stim­u­lat­ing and engag­ing, and hid­ing dis­cus­sion that the user finds off-topic, spammy, exces­sively or insuf­fi­ciently detailed, etc.

The beauty of per­son­al­iza­tion is its flex­i­bil­ity. It does not force a par­tic­u­lar style of dis­cus­sion. If the user enjoys:

  • heated dis­cus­sion back-and-forth discussion,
  • calm dis­cus­sion with well-reasoned but con­cise arguments,
  • in-depth aca­d­e­mic discourse,
  • tabloid-like ad-hominem, or
  • trolling and hate speech

then the user gets what they want.

Addi­tion­ally, per­son­al­iza­tion is adapted on a per-topic basis. One par­tic­u­lar user might enjoy a heated dis­cus­sion about abor­tion, a calm well-reasoned dis­cus­sion about NoSQL data­bases, and low-brow dis­cus­sion about celebrity romance. Per-topic per­son­al­iza­tion can sat­ify all these user needs.

I can dis­cuss more details about this approach, includ­ing how to:

  • cap­ture per­son­al­iza­tion infor­ma­tion through user inter­ac­tion with the dis­cus­sion board.
  • incor­po­rate atomic commenting.
  • fed­er­ate dis­cus­sion across mul­ti­ple sites and lib­er­ate dis­cus­sion from a sin­gle site.

I can also dis­cuss pos­si­ble objec­tions to per­son­al­iza­tion, and my response to them.
Due to space lim­i­ta­tions (500 words), I omit these details for now, and focus on per­son­al­iza­tion, which I believe addresses the core prob­lems of tra­di­tional dis­cus­sion systems.

  • Mike Altar­riba

    I’ve spent a lot of time par­tic­i­pat­ing in online dis­cus­sions, going all the way back to the elec­tronic bul­letin board days of the early 1980s, and there are some things I’ve noticed:

    * audi­ence mat­ters — Usenet (the global text-based news and dis­cus­sion sys­tem which pre­ceded the World Wide Web) was, at one time, inhab­ited almost exclu­sively by acad­e­mia, stu­dents, some busi­ness­peo­ple, and other pro­fes­sion­als. Every Fall, a new crop of stu­dents would start show­ing up, and would have to be accul­tur­ated as to the ‘neti­quette’ of the Usenet world. Then, in Sep­tem­ber 1993, AOL (a pub­lic ISP) added direct access to Usenet from their sys­tem… and the “Eter­nal Sep­tem­ber” began, so named because now Usenet received not a trickle of new indi­vid­u­als once a year, but a con­tin­u­ous flood… and the Sig­nal to Noise ratio dropped pre­cip­i­tously, and stayed that way.

    * text com­mu­ni­ca­tion makes mis­un­der­stand­ing and anti-social behav­ior *easy*, and cir­cum­vents the social reg­u­la­tory mech­a­nisms we have when it comes to face to face interaction

    * if indi­vid­u­als have their input fil­tered based on their own sub­ject mat­ter / view­point pref­er­ences, this will drive their world view such that their posi­tions and beliefs will become even more polar­ized, while simul­ta­ne­ously giv­ing them the false sense that they hold views which are shared by, and sup­ported by the majority 

    I agree that “The core value of a dis­cus­sion sys­tem is to encour­age stim­u­lat­ing and engag­ing discussion.”

    I don’t, how­ever, think that per­son­al­iza­tion will help, because it does not address the effects of dis­rup­tive peo­ple or dis­rup­tive posts… the fact that my choices mean that *I* don’t see a par­tic­u­lar poster or post doesn’t change the fact that said poster / posts are still there, and still affect­ing the char­ac­ter of the online community.

    And that, I think, points us to what is needed: A sys­tem which fos­ters a stim­u­lat­ing, engag­ing *com­mu­nity* which will in turn fos­ter stim­u­lat­ing and engag­ing *discussion*. 

    To do that, we need a sys­tem which pro­vides some sub­sti­tute for the social checks and bal­ances we have inher­ent in face to face, com­mu­nity inter­ac­tion, the checks and bal­ances we’ve evolved over mil­len­nia of social and cul­tural evo­lu­tion. I think the work being done on rep­u­ta­tion economies and trust economies points us in a pro­duc­tive direction.

    Since I’ve already been long-winded enough, I’ll stop here, but will be glad to con­tinue this if there is interest.

  • Alexan­dre Passos

    Mike, I think there is a mis­take in your crit­i­cism of per­son­al­iza­tion. You point out that “ the fact that my choices mean that *I* don’t see a par­tic­u­lar poster or post doesn’t change the fact that said poster / posts are still there, and still affect­ing the char­ac­ter of the online community.”.

    This, how­ever, assumes that there is such a thing as _the_ char­ac­ter of the online com­mu­nity. If ultra-personalization as in this post is done, each per­son will expe­ri­ence one part of a con­tin­uüm of dif­fer­ent com­mu­ni­ties. So only troll-intersted peo­ple and troll-like peo­ple would ever se the trolls, ide­ally, and hence they would harm no one.

  • http://metaoptimize.com Joseph Turian

    To respond to your objec­tions:
     
    “audi­ence mat­ters” Agreed. Ide­ally, per­son­al­iza­tion would auto­mat­i­cally find the right com­mu­nity for you, and fil­ter out the trolls.

    text com­mu­ni­ca­tion makes mis­un­der­stand­ing and anti-social behav­ior easy” Peo­ple will learn quickly how to mod­u­late their tone if the sys­tem buries their com­ments or if angry writ­ing leads to stu­pid responses.

    if indi­vid­u­als have their input fil­tered based on their own
    sub­ject mat­ter / view­point pref­er­ences, this will drive their world view
    such that their posi­tions and beliefs will become even more polar­ized” Okay, so I believe this is the most com­mon mis­con­cep­tion about my pro­posal. I’m going to address this in an upcom­ing post. But the short response is: Per­son­al­iza­tion of dis­cus­sion doesn’t fil­ter to give you a warm fuzzy feel­ing in your tummy that every­thing is right in the world. Per­son­al­iza­tion of dis­cus­sion gives you stim­u­lat­ing dis­cus­sion, which often-times means oppos­ing view­points. It should also tie in dis­cus­sions on adja­cent top­ics, to give a broader per­spec­tive on the con­text, assum­ing you are the sort of per­son who doesn’t like to have blind­ers on.

    the fact that my choices mean that I don’t see a par­tic­u­lar poster or
    post doesn’t change the fact that said poster / posts are still there,
    and still affect­ing the char­ac­ter of the online com­mu­nity.” This is an inter­est­ing argu­ment about ambi­ent dis­cus­sion. I’d like to explore it more. Can you give some exam­ples? It seems to me that if some­thing is indi­rectly stim­u­lat­ing dis­cus­sion that is rel­e­vant to me, I should get shown it.

  • Anony­mous

    I don’t know about this.

    Per­son­al­iza­tion of com­ments streams can also lead to “I only see what I like” type of stream.

  • http://metaoptimize.com Joseph Turian

    To quote my response to Mike Altarriba:

    I believe this is the most com­mon mis­con­cep­tion about my
    pro­posal. I’m going to address this in an upcom­ing post. But the short
    response is: Per­son­al­iza­tion of dis­cus­sion doesn’t fil­ter to give you a
    warm fuzzy feel­ing in your tummy that every­thing is right in the world.
    Per­son­al­iza­tion of dis­cus­sion gives you stim­u­lat­ing dis­cus­sion, which
    often-times means oppos­ing view­points. It should also tie in dis­cus­sions
    on adja­cent top­ics, to give a broader per­spec­tive on the con­text,
    assum­ing you are the sort of per­son who doesn’t like to have blind­ers
    on.

  • http://twitter.com/newsery10/status/75245062996508672 newsery10

    Dis­cus­sion 2.0: Per­son­al­iza­tion — http://bit.ly/is9arj — [Hacker News FH]

  • http://twitter.com/hnfirehose/status/75245448461426688 HN Fire­hose

    Dis­cus­sion 2.0: Per­son­al­iza­tion: http://bit.ly/m1UX45

  • http://twitter.com/bundleofthought/status/75255747352596481 Thought Pro­vok­ing

    Dis­cus­sion 2.0: Per­son­al­iza­tion: Com­ments http://digfoc.us/my89Nd

  • http://twitter.com/playlablondon/status/75282084758945793 play­lab LONDON

    re-thinking our assump­tions about com­ment­ing — dis­cus­sion 2.0: http://bit.ly/jFgmSY

  • Olivier Breuleux

    My imme­di­ate con­cern was, too, that per­son­al­iza­tion would lead to an “I only see what I like” kind of feed. I see now that this is not what you meant, but then I’m con­fused about what exactly you do mean.If per­son­al­iza­tion can show me dis­cus­sions about top­ics I like, that would indeed be useful. However, I am not sure about fil­ter­ing the type of dis­course: from your post I under­stand that per­son­al­iza­tion might skew somebody’s feed towards either a flame-war style of dis­cus­sion or towards a more rea­soned aca­d­e­mic style. Since argu­ments from both sides of any issue are con­sid­er­ably worse in the for­mer case than the lat­ter, I am afraid this might lead to an impov­er­ish­ment of dis­course, with facts, evi­dence and good argu­ments fail­ing to reach a large seg­ment of the pop­u­la­tion. To com­pli­cate mat­ters even more, there is some­times a divide where heated dis­cus­sions lean on one side and rea­soned dis­cus­sions lean on the other.In my opin­ion, what a great dis­cus­sion sys­tem needs is first and fore­most a way to pro­mote facts and sink mis­in­for­ma­tion. I would also like hate speech and ad hominem to be penal­ized for every­one. While some peo­ple might enjoy hate speech, I do not think soci­ety ben­e­fits from giv­ing them what they want. A great dis­cus­sion sys­tem needs to improve peo­ple, rather than sim­ply fit their biases, and even though I know the lat­ter is not what you intend, it seems very dif­fi­cult to avoid it. The global effects of a dynamic sys­tem can be very dif­fi­cult to pre­dict, and by try­ing to make a per­son­al­ized expe­ri­ence, you might inad­ver­tently end up dis­trib­ut­ing less qual­ity to less peo­ple — I sus­pect this is an attrac­tor of per­son­al­iza­tion when it is applied to debate, though you might disagree.

  • Olivier Breuleux

    Sorry, I thought these would make para­graphs :(

  • http://profiles.google.com/oliver.mitevski Oliver Mitevski

    It dis­turbed me that the TED talk about “fil­ter bub­bles” immensely  overem­pha­sized the flaws of per­son­al­iza­tion on the web. Nevertheless it’s a tough prob­lem.
    One idea could reuse the aard­vark idea. In my per­son­al­ized pro­file, the sys­tem could serve me with ques­tions that I could pos­si­bly answer or dis­cus­sions I could join, based on my cur­rent answer­ing record. This is actively ask­ing me to answer/join, by send­ing my an email once a week for example.This way it would be more engag­ing for professionals/experts/scientists, because they will be served with pro­vok­ing high-level questions/discussions, and the eas­ier ques­tions would be served to less expe­ri­enced so they have a low entry bar­rier, and find it more engag­ing. Occa­sion­ally the sys­tem has to explore in order to exploit more effec­tively (rein­force­ment learn­ing). Basi­cally the experts will be served with some new­bie ques­tions to test them and the new­bies will be asked to par­tic­i­pate into some more seri­ous discussions.Also peo­ple have var­i­ous behav­ior pat­terns. Some have inter­ests into var­i­ous top­ics, but cover them super­fi­cially, oth­ers only one but cover it in full depth. Bet­ter to com­pute exper­tise with respect to a topic. Any­way more min­ing of pro­file behav­iors is needed.

  • http://twitter.com/openinnovation3/status/75475800534679552 openin­no­va­tion

    Dis­cus­sion 2.0: Per­son­al­iza­tion – MetaOp­ti­mize: http://bit.ly/kXTlxv

  • http://twitter.com/omitevski/status/75536759689191425 Oliver Mitevski

    Dis­cus­sion 2.0 Per­son­al­iza­tion to the res­cue! http://bit.ly/jbZPbd

  • Brian Van­den­berg

    It would be nice if some­thing like this could become a real­ity, but the sad truth of it is that stack exchange-like sites thrive pre­cisely because there’s a car­rot dan­gled in front of people’s faces: you will become more respected, and peo­ple will care more about what you say, if your ‘karma’ score, or your ‘rep­u­ta­tion score’ are higher.

    This has been proven time and time again in many fields, among them my favorite exam­ple is the story of Vir­ginia Apgar (founder of the Apgar score): prior to pub­lish­ing about the Apgar score, the rate of sur­viv­ing live births was abysmally low, but as soon as doc­tors had some­thing to mea­sure them­selves by the sur­vival rate sud­denly sky-rocketed — if for no other rea­son than because they had some­thing to brag about while play­ing golf.

  • http://twitter.com/htayyar/status/113254951765028865 Hasan Tay­yar BEŞİK

    Dis­cus­sion 2.0: Per­son­al­iza­tion – MetaOp­ti­mize http://j.mp/oYgmow

  • http://twitter.com/appliedai/status/132810135516098560 Applied AI

    RT @turian: Dis­cus­sion 2.0: Per­son­al­iza­tion http://t.co/KjHrC6Lw

blog comments powered by Disqus