Democracy’s Soft Underbelly

Let me explain.

It’s now well-documented that outside forces took advantage of social media platforms to spread rumors in order to swing the 2016 presidential election. Journalists digging into the story are looking more closely at the tools used to purchase advertising that helped amplify these rumors, and are horrified by what they are discovering.

Last week Pro Publica discovered you can target “jew haters” and BuzzFeed News found that on Google you could target phrases such as “blacks ruin everything”

In the days of print, each advertisement was reviewed by multiple people from both the organization that bought the ad and the publication that ran it. Extreme care was taken to make sure the advertising complemented the editorial and the message was the right fit for the audience, not only to maximize effectiveness but also to avoid instances such as the one below.

Despite careful review, print ad placements sometimes backfire

Online advertising is a delicate balance between scale and quality. The dream is to serve a perfectly targeted ad to as many people as possible. But because of the scale, it is impossible to manually review each and every ad creative for quality and fit. In the online world, people “optimize” and let the algorithms do the work.

While at Yahoo, I met with an advertiser who wanted to learn about our behavioral targeting options. I was working with a team that was thinking about exposing detailed facets of the massive Yahoo audience that would help advertisers reach very specific segments. When I walked into the room, the client had a spreadsheet he was using to allocate his million-dollar budget. After asking a few questions about his goals, I proposed a few very targeted criteria to build his target audience. Unfortunately, he grew frustrated because the total audience was too small and we were going to have to run hundreds of queries to build up the reach he needed. He didn’t have the time to continue the exercise nor appetite to keep track of all the data to show ROI to his client. The meeting wrapped up with four very broad buckets into which he poured roughly $250k each and called it a day.

He couldn’t be bothered with the details.

This is the state of online advertising today. The tools available to reach massive scale are even more sophisticated but to do it right, with quality, requires manual oversight. Ad units can be configured to dynamically swap out ad copy and assets depending on the target audience, which can also be built algorithmically. Ad spend adjusts automatically and APIs monitor trending keywords to take early advantage of trending topics and get broad reach on the cheap.

“Programmatic Advertising” is a blanket term for techniques used to automatically generate thousands of “personalized” ads at massive scale. Because it’s automated, generating ad copy variants and target segments is inexpensive. The downside is that quality suffers if you take out the human element, leaving the robots to mind the store.

This slideshow requires JavaScript.

There are many examples that show us that resulting matches still need regular review. As long as ad matching algorithms such as Google’s and Facebook’s remain black boxes, a regular human review is necessary to prevent the unexpected.

Which brings me back to what we’re learning today. Last week Facebook shared that ad placements made to “amplify divisive messages” were used to influence the 2016 elections. We are slowly uncovering the extent of information warfare that uses social media platforms to weaponize fake news. Using programmatic advertising to draw attention to and amplify these campaigns is a natural extension.

More careful review of editorial content posted to social networks is important to verify facts and prevent the spread of “fake news.” On the flip side, advertising platforms need review as well because ad targeting is also rife with repugnant audience segments automatically suggested by the algorithms. As they should, Facebook and Google have both said that more rigorous review is on the way but there will always be the tension of profit motives to discourage too rigorous a lens.

In the late-90’s, the movie Wag the Dog spins a tale of how a Washington “spin doctor” (Robert De Niro) hires top Hollywood producer (Dustin Hoffman) to engineer public opinion. The film made light of the gullible public but there was a broader, cynical message about how media (and the press) can be used to manipulate public opinion.

Today we are seeing this same scenario played out, but instead of manipulating public opinion through TV and Hollywood, public opinion is bought and sold using social networks and online advertising.

Google DeepMind Plays Go

There is a Challenge Match taking place in Seoul between Google’s DeepMind AlphaGo computer program vs. 9 dan professional Lee Sedol (9 dan is the highest rank). Most of the engineers at SmartNews have a background in machine learning and are following the matches closely on a dedicated internal Slack channel.

The YouTube coverage is very good with professional English commentary from Michael Redmond, the first Western Go player to reach 9 dan. Go is a fascinating game and Michael’s commentary is quite good and easy to understand even for beginners like me.

The first two matches went to Google and it looks like history is being made. I’ve embedded videos for the upcoming matches as well.

Update – AlphaGo wins in three.

Update – Lee Sedol wins match four!

Match Five

Fun with Google Trends Real Time

Google Trends announced last week that they’ve upgraded their service to be real time. Using their tools, I created a dashboard so you can quickly see who’s trending in Google Search for the past 7 days. If I did this right, this page should continually update.

I’ll manually add/remove names as the list of candidates change.

Democratic Party

Republican Party*

*There are too many in the field to fit on Google’s graph (Google Trends only takes up to five terms to compare). I took the top five announced candidates in the polls.

Google Maps Gallery

10 years ago, when I was looking for a place to live, I had three maps to help me zero in on where to look. I was concerned with schools so I had a map from greatschools.org along with a school district map showing which houses served which schools. I then had a real estate map from realtor.com that showed the price of houses in the area. Back then the wish was to overlay the two maps on top of each other and, indeed, some of the original mashups which kicked off the Web 2.0 movement were driven by these types of demands.

Since then, the Google Maps teams has been busy pulling in all sorts of layers together and have gathered them all together into their Google Maps Gallery which launched today. There’s a load of things to get lost in (including the overlay of San Francisco in 1938 shown above).

Read their blog post to learn more.

Technology as Connective Tissue

Two tear-jerker videos illustrate the power of technology to connect over distance. Watch and marvel the world we live in. Happy Thanksgiving everyone.

Google Search services connect two old friends across political boundaries. This video was put together by the Google India team.

Skype connects two girls on opposite ends of the earth who share a disability, part of Skype’s Stay Together campaign.

Google’s Android Dream at Scale – Moto X

Moto X

Wired’s man on the ground at Google, Steven Levy, has an in-depth look at the turnaround story of the Motorola Mobility team purchased by Google for $12.5 billion two years ago and how they produced a phone which, on the eve on iPhone’s expected upgrade in September, is currently the talk of the Valley. As far as specs, it’s not running the latest and greatest but that’s just fine as the target audience is not the high end gadget freak, they are elevating the bar so that the masses can experience the fully integrated Google vision.

But the defining feature of the Moto X is it’s a virtual ear, always straining to hear its owner’s voice say three magic words that will rouse it to action: “Okay, Google Now.”

Here is a phone that is always waiting, ready to spring into action even faster than Apple’s Siri. Sure it’s always listening to you but in return you get a phone that can predict your needs with Google Now-enabled prescience. All that stuff that we technologists all dream of but ultimately fail at because of competing standards, incompatible platforms, and flaky APIs are now possible because Google owns not only stress-tested services in the cloud but also the end device.

  • an instant signal when you walk in a restaurant that starts a stream menus and reviews
  • warn you to end a meeting because it knows that traffic is so snarled, you might not make your next one in time
  • Only fools don’t protect their phones with a password, but it’s a pain in the neck to punch it in a few hundred times a day. Motorola plans to ease that pain (though not available at launch) by selling plastic tokens that can clip onto clothing—if the tab is within a few feet if the Moto X, no password necessary. (The tokens use NFC technology, built into the phones.) The Moto X will also let you set up password-free “safe zones” like your car.

These are just a few examples quoted in Levy’s piece. A few more were discovered by my colleague who is testing out a demo unit,

  • the phone uses its GPS to determine when you might be behind the wheel of a car. Assuming that you are, this function can read aloud incoming text messages automatically. It can also send an auto-reply in this situation.
  • Meeting mode works off of your Calendar events. When the phone sees you’re in a meeting, it can automatically silence the handset. You can allow Meeting mode to ring the phone or auto-text replies to favorite contacts or if anyone calls twice in a five-minute period.
  • You link your phone and your Chrome browser through an extension so you can get caller or text information when on your computer. No need to pick up your phone for that data and you can also choose not to pick up the phone if you don’t want to take the call. You can also reply to text messages from your computer browser.

motorolaconnect

Feature List for an RSS Reader

With the announcement of the sunsetting (never did like that word) of Google Reader, a discussion was kicked off at work over what features would make up an ideal RSS reader. Everyone at GigaOM is a voracious reader so we like to compare information processing tools and techniques like foodies discuss recipes.

rss-buttons

Here’s my short list:

  • Must be able to import an OPML file. The easiest way to get started is to load up your existing collection of feeds.
  • Must export OPML. Never trust a platform that doesn’t support data portability.
  • Must keep track of what you’ve read.
  • Must have a mobile version that syncs what you’ve read with on the desktop, mobile, or anywhere else
  • Must support pubsubhubub so news is pushed and realtime if the feed supports it.
  • Must be able to browse by feed or as an aggregated, reverse-chron sorted river of news
  • Must support browsing by headline, excerpt, or full-text
  • Must support rich media so the reader can be used to browse video, podcasts, and photo feeds. Bonus points if you can output a photo feed as a screensaver.

Then there are the extra features are what would put one reader above others

  • Provide search across all feeds. This is your slice of the best of the internet after all.
  • Add the ability to star or otherwise mark items for simple re-tweet behavior. Let people publish a feed of these curated items so others can follow your information exhaust. Even better is to re-create the “share with note” feature in Google Reader and you’ve got a light-weight tumblr network.
  • Add the ability to follow other people and add their feed bundles to your collection. This was the single best feature of Google Reader and the one that, when taken away, killed off the future of the product.
  • Decay. Add a natural decay to feeds that do not get a lot of your attention. Provide a bookmarklet that lets you grab and add feeds as you find interesting posts across the internet but feel safe in the fact that if you let a feeds’ post go unread, that the feed itself will eventually drop off your main view, keeping things clean and focused.
  • In the day and age of Twitter & Facebook, have a pre-set filter that reads your social feeds and parses out all the links you add and puts them into a folder which you can search across or curate & share back out.

Finally, there is the uber-geeky-cool feature that I built with the MyBlogLog team, the Interest Engine. The vision was that you would pipe all your feeds through the reader and the tags on all those feeds and shares would feed the algorithm to improve what bubbles up in your aggregated newsfeed. If you subscribe to a bunch of blogs about “fly fishing,” use that as a signal and focus posts from other, more generic feeds on your interests so that if a story about Fly Fishing flows across your New York Times feed, it gets higher placement.

So that’s my list of MVP features & nice to have differentiators.  Did I miss any?

UPDATE:

Some choice words from Chris Wetherell, one of the original engineers on Google Reader, on the effervescent business opportunity of the GReader community.

Dave Winer shares his thoughts on how he would build RSS anew. Centralized OPML profiles (as were offered by GReader) are key.