Deep-learning, Dark Social, and Internet Freedom

Freedom on the Net 2014Making Sense:

Signal and relevance are tied to our ability to see and observe; what happens when the context in which we operate is very much part of the narrative, or, in some cases dominates it?

  • Why Are Some Cultures More Individualistic Than Others? NYTimes: For example, Americans are more likely to ignore the context, and Asians to attend to it. Show an image of a large fish swimming among other fish and seaweed fronds, and the Americans will remember the single central fish first. That’s what sticks in their minds. Japanese viewers will begin their recall with the background. They’ll also remember more about the seaweed and other objects in the scene.
  • Freedom House report: Internet freedom declined in 36 of 65 countries studied this year. The New YorkerWhat’s behind the decline in Internet freedom throughout the world? There could be several reasons for it, but the most obvious one is also somewhat mundane: especially in countries where people are just beginning to go online in large numbers, governments that restrict freedom offline—particularly authoritarian regimes—are only beginning to do the same online, too.

Making Do:

Marketers are focusing more and more on numbers. What happens when the numbers do not tell the whole story, or end up not being wholly accurate?

  • Google admits that advertisers wasted their money on more than half of internet ads. Quartz:  Google announced that 56.1% of ads served on the internet are never even “in view”—defined as being on screen for one second or more. That’s a huge number of “impressions” that cost money for advertisers, but are as pointless as a television playing to an empty room.
  • Chartbeat Shines A Light On Dark Social, Finds Missing Mobile Referrers. Marketing Land: Chartbeat’s adjustment came after it discovered that a number of highly used mobile apps have added the app identity to the user agent setting, which typically is used to identify the browser and operating system of web pages. Now the source of mobile app traffic from Buzzfeed, Pinterest, Flipboard, Chinese social network Tencent QQ and Chinese search engine Baidu, among others, is being properly attributed in Chartbeat.

Making It:

Can we have our predictive, useful, and contextual data cake and eat it too? As these stories suggest, results may vary, for now.

  • Deep-learning startup MetaMind launches with $8M from Benioff & Khosla Ventures. Venture Beat: A type of artificial intelligence, deep learning involves training systems on lots of information derived from audio, images, and other inputs, and then presenting the systems with new information and receiving inferences about it in response. […] Use cases could include extracting key signals hidden in financial analysts’ reports, or analyzing chat messages from people seeking customer support from a company.
  • Algorithm fatigue: What Evernote’s news-recommending product can tell us about privacy. Nieman Lab: For publishers, Context is a unique, if niche, way to get their content in front of engaged, professional eyeballs. What Evernote hopes to offer those users attached to those eyeballs is a productive, predictive way to work the likes of which they’ve never experienced before. […] Sinkov […] believes that if people better understood how Context works — that the data flows only one way and that there’s no money involved — they would be more open to the benefits of the product.

[image via]


Valeria is an experienced listener. She designs service and product experiences to help businesses rediscover the value of promises and its effect on relationships and culture. She is also frequent speaker at conferences and companies on a variety of topics. Book her to speak here.

, ,