Saturday 12 March 2016

Your Facebook Newsfeed is about to get more Expressive

Your Facebook newsfeed is about to get more expressive in positive way. We all are expecting dislike button for a long time but Mark Zuckerberg always said that he feels, dislike is attached with so negative feeling and thinking so he never prefer that extra feature in NewsFeed. In last Q-A round of 2015, he announced that they are working on some same kind of thing and finally that thing came out before just some days. Yes I am talking about "New Reactions Roll-out Change of Facebook"

Say "Hi" to new reactions

Personally I like Facebook all time and the main reason is, they are always courageous in change factor. I have one page for my blog, what I observed and analyzed through it, is that they are changing their algorithm more frequently.

Sammi Krug, Product Manager of Facebook wrote down very clearly in their one of the blog post, "New NewFeed and ADs algorithm is about to come and it's just because of these new reactions". They are usually using sum of all reactions as likes on post and run their old algorithm but in next level of reactions they are about to give different weighting to those reactions on which are more frequently used reactions for you. The main aim of NewsFeed algorithm is to provide more valuable content for you and using those reactions it will be very easy for them to do.

Do you really think those reactions are just for more convenient way to express user's feelings ? then you failed to understand it, these reactions lead to easy way of sentimental analysis of user's feeling towards any AD or post of page. More business analysis on ADs, Any Page Admin come to know which kind of post user are expecting from page and many more changes.

Why just Five extra reactions, Do you really need extra ?

Paradox of choice by Barry Schwartz, This book simply says Why More is Less. more choices you get, more confusion you will get with your choice. Julie Zhno, Product design director and Dacher kelther, Professor of social psychology at UC Berkely analyze all one line comment and most used stickers in comment to get out those five basic reactions. The fact is they tried other designs of those reactions in testing phase which I shown in below image.

Tested Designs for New Reactions but at last they decided to go for GIFs

What are those Lucky countries, where testing phase of any new change is happened very first ? 

As per data given on tech-crunch website, Adam Mosseri, product director of Facebook said that this time they tested new reactions in Spain and Ireland. The reason is both of them have large national users than international. ("High Modularity ?? ")

Computer based personalities Judgments on Facebook Likes beats Human Judgments(Research by Wu Youyou, Michal Kosinski and David Stillwell)

 

Knowingly and Unknowingly we judge other's personalities. How they interacts, what are the facial expressions used and many questions are running in our mind while talking to new person. Even by watching one small video-clips we start to judge strange people in video. 

In this experiment of three psychologist, they taken 86220 volunteers for Human based judgments and 70520 Facebook user's likes for Computer based judgments. while analyzing, problem is perspective of people for any person change as per relationship measurement. Solution of this problem is using various perspective of personality. 

Three perspective measurement of personality :
  1. Self-other agreement : When friend's belief about any volunteer matches volunteer's belief about themselves.
  2. Inter-judge agreement : Consistency in judgment while two friend's of any person judge him.
  3. External validity : conclusion of analysis on our dataset work for other dataset. 
Big Five personality traits :
  1. Openness : inventive/curious vs. consistent/cautious
  2. Conscientiousness : efficient/organized vs. easy-going/careless
  3. Extra-version : outgoing/energetic vs. solitary/reserved
  4. Agreeableness : friendly/compassionate vs. analytical/detached
  5. Neuroticim : sensitive/nervous vs. secure/confident

In Human based judgments, sample of 86,220 volunteers,who filled in the 100-item International Personality Item Pool (IPIP) Five-Factor Model of personality questionnaire. Computer-based personality judgments used LASSO (Least Absolute Shrinkage and Selection Operator) linear regressions with 10-fold cross-validations, so that judgments for each participant were made using models developed on a different subsample of participants and their Likes.
  

Result of Experiments : 

 

1.  Self-other agreement :

Here participant's scores is corrected with the judgments made by humans and computer models. Result of self-other agreement also depends on relationship between participant and human judge like friends, spouses, family members, cohabitants, and work colleagues. as shown in below image,Computer's average accuracy steadily grows with no of likes. on an average computer's average accuracy(0.56) beats Human's average accuracy(0.49).

 

2.  Inter-judge agreement : 

The inter-judge agreement for humans compute using a subsample of 14,410 participants judged by two friends and then average it on collection. Inter-judge agreement for computer models compute by splitting Likes into two halves and developing two separate models and then compare it. Again computer's average accuracy(0.62) beats Human's average accuracy(0.38).

3. External validity :

External validity focuses on how well a judgment predicts external criteria, such as real-life behavior, behaviorally related traits, and life outcomes. Here we compare 13 criteria and computer judgments was higher than that of human judges in 12 of the 13, as shown in below image.


 

Conclusion ::


5 more reactions is added in list with likes. Computer based judgments are absolutely going to be more accurate with reactions for Facebook NewsFeed. More personalized NewsFeed is about to come. 
  

Source of Information ::

  1. NewsFeed News by By Sammi Krug (http://newsroom.fb.com/news/2016/02/news-feed-fyi-what-the-reactions-launch-means-for-news-feed/)
  2. Interview with Product Design Director Julie Zhuo (http://www.wired.com/2016/02/facebook-reactions-totally-redesigned-like-button/)
  3. Youyou, Wu, Michal Kosinski, and David Stillwell."Computer-basedersonality judgments are more accurate than those made by humans."
 
 

1 comment:

  1. Good article! Learned a few things. They still kept sad and angry though and they have not published any research numbers.

    ReplyDelete