Jan 8, 2016

Signal and Noise on Twitter

Twitter is a great place for new and interesting information -- if you can find it!  So much of what is good on Twitter is drowned out by the noise on Twitter.  At the end of 2015, I was dissatisfied with my Twitter experience, there was too much water to wade through in the Twitter-stream. I thought about following less people, but realized I should be following the right people, or as close as I could get to that.

Others have also been trying to find how to get the Signal out of Twitter without drowning in the Noise.  My friend, Luis Suarez (@elsua) took a radical step with his Twitter activity by unfollowing everyone and then re-following select groups of people via Twitter Lists.  In a recent conversation, he said his approach is working out well, but he has noticed that he needs to edit his lists after several months.  In a different approach, 5 years ago, I looked at algorithms that would help me to figure out who to pay attention to on Twitter.  The LFR algo was great for a quick calculation, but it excluded too many fine folks -- in reducing noise well, it also often got rid of the signal too.

When we have good information, decisions are easy to make.  When we have very little, or no information, we look to others and see what they are doing while in situations similar to ours.  It is easy to select who to follow from people we know and have experience with.  It is also easy to add a person to the Following list when a friend/colleague we trust, excitedly proclaims a person's expertise in a field/topic we are interested in.  But what do we do when we run across someone on Twitter who is in our field, and has tweeted something that caught our eye, but we know nothing about him/her?  Sure we can read their short Twitter bio, or look at the number of their followers, or even read some recent tweets, but do these reveal an accurate picture of the person?  No, not in my experience.  I have followed many, after a brief review of the above info, just to find their tweet stream full of noise.  I then unfollow them and wish I had picked better.

So, I am going to follow Luis' lead and unfollow everyone in my personal Twitter account (@valdiskrebs).  Most of the people I currently follow I will follow back via one of three Twitter Lists I have created.

  1. The first list is purely algorithm-driven. This new algorithm is an improvement over the LFR algo I have used in the past. It attempts to find those Tweeps who have more Signal then Noise in their Twitter activity.
  2. The second is all human choice.  These are people I know, have been following for years, and have interacted with either online, or offline, or both.
  3. The third is a little of both -- algorithm and human choice.

Although I use AI -- artificial intelligence in my network analysis work, I am a big believer in IA -- intelligent assistance.  Algorithms do not give me the final, or only answer, they help him make the best decision. Algorithms help organize my data and information and reveal patterns I may not have been considering. In the end, based on everything before me, I decide. That is why I plan to keep adapting these lists, and also adapting the new algorithm. 

For those that have followed me for a long time, please do not take this mass unfollow as a personal rejection -- it is not.  If we know each other, I am following you back on the second list mentioned above.  Those that will unfollow me, because, and only because, I have unfollowed them, well... see ya!  While I am still refining this approach, I will keep the three lists -- X, Y and Z -- private.  In order to have conversations (direct messages) with those who follow me, I will open up the Twitter access for Direct Messages to everyone.  I hope this will not result in a lot of spam from Twitterbots and other mass marketers.

If the algorithm works out well I will offer it to my clients, if not, I will be smarter about many things.  Some time in the future I will update everyone on my progress.  New year, new thinking.

Happy 2016 to everyone!

Update Feb 6, 2016: There is news that Twitter will change how they display everyone's timeline (the people you are Following).  If you do not want to view Twitter's new algorithmic timeline (we do not know how they will filter/adjust it at this point), just create a Twitter List of your current following group and you will still control who you see.  In your Twitter client, chose to display the List instead of your standard timeline.

Update August 12, 2016: After experimenting with my personal account I learned quite a bit.  So, I have decided to start following a small group on my personal account that old-fashioned way and switch my business account to Following totally/only by lists.  It makes much more sense to have focused following lists in business than in your personal life (IMHO)!  So, I have created 4 focused Twitter lists to follow about 600-800 people via my business account.  These lists are both algorithmically and humanly curated.  I adjusted my algorithms as I went through the learning process with my personal account -- they are greatly improved from where I started! I think this new approach will work well, but you never know what you will learn!  Will report back at the end of the year.


  1. Hi Valdis, what a brilliant blog post! Thanks for sharing it along and for the link love! Much appreciated. I'm glad to see other folks continue to question how they may good use of Twitter in order to get the most out of this social networking tool and I will be looking forward, down the line, about how your initiative works out and whether the algo works out all right, or whether it would require some tweaking further along...

    In my own experiment with unfollowing everyone, and as you well noted above, I'm enjoying it quite a bit as Twitter has transformed itself a little bit into a Slack space with multiple channels (i.e. Lists) where I get to interact and engage with folks based on a proximity basis and topics of interest, making it much more meaningful than the constant stream of tweets with very little context and sense added into it. It's transformed entirely how I get to use Twitter nowadays and, despite having started using it a few months back, I'm still enjoying it as much as ever, to the point where I will carry on with it this year as well.

    Yes, you are right as well in that every few months I need to go through the Lists to continue to fine tune them, but, over time, I have found it's a natural process of tuning your own interactions, conversations and interests. I mean, people who I used to converse and talk with, and learn some more from / with, I'm finding out those affinities are no longer there as we may have both followed different pathways, so instead of lingering around I decided to 'just let them go', just like they did, and move on with the times into better pastures. I guess it's how we build relationships, whether offline and / or offline, so trying to stick around to see if it would click again doesn't make much sense anymore to me and I guess to them either, since I haven't heard from those who have dropped off the lists in last few weeks.

    The end result is that each of the public Twitter Lists becomes more focused, intentional, purposeful on what I would want to achieve, i.e. whether to collaborate, cooperate or learn from others and, like I said, so far I'm quite liking that approach, for sure!

    I'll be curious to find out plenty more how that blend of the algo and the intelligent assistance would work out for you and whether you'll see any differences from precious uses of the tool and will be looking forward to exchanging some more notes on what we both learn. But, again, what I'm enjoying the most about this initiative is how more and more people are starting to question the use they do of certain social tools to justify their own participation and involvement with each of them. And that's a good thing, from what I can tell :)

  2. That's great Luis! We are in violent agreement! ;-)

    Also, thanks for recommending Tweetbot to me -- it makes following Lists very easy/useful. So much better than the standard Twitter app or web interface.

    1. Hi Valdis, you are most welcome! Happy to help where I can :) And, yes!, Tweetbot (v4. specially on iOS or v2.x on the Mac) make Twitter really worth while, not just for Lists, but in general. I know it misses some of the new functionality put together by Twitter like Polls, Hearts, Moment, and a couple of other things, but if you don't really mind about those 'new' capabilities and would want to have a clean experience without ads, Tweetbot is the way to go, still. Totally worth while the investment! Glad you are enjoying it as well :)

  3. Hi Valdis. I was wondering if this might have been inspired by Luis. I hope you follow up this post with how the lists are working for you. I'm very interested in this tension between AI and IA. Did you read John Markoff's latest, "Machines of Loving Grace"? That's the major theme in the book.

    One quick point that is worth making. This still doesn't solve the problem...for me at least. I have two modes that I go into on social networks: topic mode and relationship mode. Relationship mode is when I want to show a little love to specific sets of people or simple tune into what they're focusing on because, holistically speaking, they share interesting stuff. Topic mode is when I want to tune into a particular topic, and Twitter's not really that good for that, aside from hashtags, which can be pretty spotty in terms of consistency. The key is either really good content-sensitive algorithms or helping people to better tag their own stuff (or some combo). That's why the new "Collections" feature on Google+ has proven interesting. Still lots to improve there, but I think it's moving in the right direction.

    1. Hi Gideon, very interesting description of your use of Twitter for those two different modes. Allow me to ask a follow-up question, if I may: have you considered a third mode where it's a combination of both topic and relationship? Somehow I keep getting dragged back into this third mode and somehow I keep finding it much more focused and purposeful to help me make sense of Twitter overall. Did you ever consider that blend of modes? I'd be curious to find out ... Thanks!

    2. I guess in practice, Luis, it's always somewhat of a blend of the two. But the problem I have with Twitter lists is that they still don't allow me to really filter by you and, say, the topic of collaboration. To Valdis' point below, you'd have to do some sort of "content sensing" to be able to apply that filter. So, on Twitter, ultimately, it's about following a person even if you use a list to track their tweets. Sometimes you have people who stay very much on one topic and so a list dedicated to that topic maps quite well to what they are tweeting about, but more often than not, their tweets will have nothing to do with the topic of the list that I just happened to assign to that person. So ultimately, I can't really seem to blend the two very well on Twitter.

      It's different on Google+ with the new Collections feature. For example, there's a guy who's senior on the G+ development team. I'm not all that interested in most of his posts there, but he has a collection where he announces changes to the G+ platform. I do care about those and so I've set G+ to notify me when he publishes something in that collection. Super powerful and quite intuitive and easy. That's what I'm talking about.

    3. Hi Gideon, thanks very much for the follow-up and for the feedback! Hummm, I am thinking that perhaps that's the beauty of Twitter, after all, the messy, chaotic, unfiltered nature of our tweets where we get to know and find out about people's interests and passions in a multi-layer capability. After all, getting to know about someone in just on single aspect is probably not much of a complete picture of who the person is.

      And the way you describe the use of Collections in G+ sounds pretty close to how one would make use of hashtags in Twitter to keep up with those conversations, without perhaps the marketing and broadcasting capability most people seem to employ when 'abusing' hashtags themselves. But there is something in them that still makes Twitter worth while the effort, as well as following people (through lists or else) to get a better picture of the person as a whole vs. just one of their interests ...

      Mind you, it would be really cool if Twitter would do a bit of filtering for us and from all of the tweets from a single individual or individuals select and curate the ones related to certain topics for us to digest accordingly. That wouldn't be too bad, after all, in my opinion, don't you think? :)

  4. Yes Gideon, I agree withg the two modes. In the topic mode I may have enough info to make a decision, or I may not. That is why there are two lists for this mode. In the relationship mode I do have enough info, I know these people, and therefore that list is easy. The algorithm only comes in to play when I do not have enough info, or I suspect the info I have. I do no "content sensing" in my algorithm. I think we basically are in agreement. :-)

  5. I am waiting to see how this develops... I don't have the background to assess the algorithmic component, however I wonder whether you three lists align with Harold Jarche's PKM ideas. Seems to fit his seek-sense-share construction which I was reminded of in Harold's recent blog post http://jarche.com/2016/01/the-core-competency/

  6. I've been using lists for a while (although I haven't made the leap to unfollowing... but I rarely read the main feed.) I find the lists to be a much more enriching experience. Have you hooked up your Twitter lists to Flipboard? It's my favorite way to go through them - easy to scan and interact.

    I'd love to get my hands on your algo - I find the list building process really tedious... and then they get dated as people's interests stray from the topic that I found initially useful. I've been building them manually using Little Bird lately. But again... super time-consuming.

  7. Great post Valdis and thanks for sharing. I must admit I've gone a bit broader than three lists. Whilst I've only a few public ones which pretty much contain between them everyone I 'follow' so to speak, I've also usually got around 15-20 on the go. Some I check in all the time to, but most I tune in as and when I need to hone into that specific topic, group, collection etc. A bit like channels on the TV if you like.

    Just wondering what you use to setup your algorithm list?