AI is only human

I’m so glad that The New York Times ran this op-ed (Artificial Intelligence’s White Guy Problem) about the inherent biases in Artificial Intelligence algorithms. Popular culture and much media coverage of AI tends to mysticize how it works, neglecting to point out that any machine learning algorithm is only going to be as good as the training set that goes into its creation.

Delip Rao, a machine learning consultant, thinks long and hard about the bias problem. He recently gave a fascinating talk at a machine learning meetup where he implored a room of machine learning engineers to be vigilant in making sure their algorithms were not encoding any hidden bias.

The slides from his talk are posted online but Delip’s final takeaway lessons have stuck with me and are good to keep in mind whenever you read stories of algorithms taking on a mind of their own.

Delip Rao takeaways

It is still very early days and many embarrassing mistakes have been made and more will be made in the future. Our assumption should be that every automated system is fallible and that each mistake is an opportunity to make things better (both ourselves and the algorithm) and should not be an indictment of the technology.

 

Bounce Rates on News Sites

I was asked by someone about typical bounce rates on news sites. Keeping with my rule that an email response sent to more than a few people is better served if it lives out on the web where it can be discovered and referred to with a link rather than locked away in an email archive, I decided to share what I wrote.

Bounce Rate is defined as, “the percentage of visitors who enter the site and then leave (“bounce”) rather than continuing on to view other pages within the same site.” Generally bouce rate more useful when you’re measuring the drop-off of visitors to an e-commerce site where you are guiding people along on a “funnel” towards a transaction or other anticipated result.

This makes it a less useful metric when measuring performance on a blog or news site where typically you serve what the reader came for on their first pageview. That said, here are some things to look at and how you can use bounce rates as an effective metric.

News site bounce rates are all over the board and I’ve seen rates anywhere between 30 – 80%. The important thing is to segment your audience into populations that have different experiences so you can see which levers are impacting your bounce rate.

Are mobile users bouncing more than desktop? Maybe your mobile website is slow.

Are users on a particular browser bouncing higher? Check your site load times.

What about regional differences? If your international readers bounce rate is higher check with Ad Ops, maybe the remnant inventory served is impacting performance.

Another cause could be click-baity headlines. Take a look at your stories that have a high bounce rate and see if the headline is saying one thing but the post another. Do less of that.

Look at your largest referral sources. Are Twitter readers bouncing higher than Facebook? Maybe your social sharing unit needs work.

One other source of bounce rate is when a reader just leaves the page open and the session eventually times out. A quick hack to record an additional event that “tricks” analytics to preventing a timed out bounce is a “toaster” which you often see appear offering a follow on page as the reader scrolls down a page. This will improve your bounce rate but I would argue that it provides little value except as yet another link to follow and may distract you from understanding your reader’s true behavior.

Finally, look at the difference between new and returning visitors. If returning visitors are bouncing, get rid of pop-ups to subscribe or other CTAs that get in the way of reading a page.

All that said, bounce rates are not that helpful to measuring a news site’s performance. Most news sites are designed to give you what you need on one page. A visitor who follows a link to a story and leaves is the definition of someone who bounces. Bounce Rates were invented to measure conversations on Anne-commerce sites funnel. For news sites you’re better off optimizing for time spent, pages/session and conversion to subscribers.

Have anything to add? Please leave your thoughts below.

Kanye being Kanye

Kanye West is a master at manipulating mass media to his advantage. Say what you will about his stunts such as renting out Madison Square Garden, his legendary twitter rants, or his jab at Taylor Swift at the VMAs, Kanye knows how to get attention. He was in top form today, using the daytime television show, Ellen, to launch into an epic rant about media, popular culture, and what it takes to change the status quo.

Social Media and the Spratly Islands

The tension around the contested Spratly Islands in the South China Sea continues to ratchet up and forces nations in the region to take sides. One of my favorite books about an earlier crisis, the Cuban Missile crisis, is Thirteen Days, which chronicled how JFK navigated his way to a peaceful resolution of the situation. My impression from the book was that we were able to walk Cuba and Russia back away from the ledge because President Kennedy was able to give Russia and Cuba room and allowed them to save face.

This will be difficult today where there isn’t such room to maneuver, where the geopolitical standoff is taking place in full view of the world and our tightly bound 24/7, interconnected networks.

But this new age brings danger as well. Such immediate transparency increases pressure on governments to respond to developments that may have been handled deliberately and privately in the past. For example, both Beijing and Washington made public statements within a day of the exposure of China’s missile deployment to Woody Island. In contrast, during the much more dangerous Cuban Missile Crisis, President Kennedy had six days to plan a response before his first public statement.

Transparency may undermine stability in a brewing confrontation, as each move and countermove is broadcast worldwide. Thus, transparency may undermine stability in a brewing confrontation, as each move and countermove is broadcast worldwide. Transforming global standoffs into spectator sports will increase public pressure on leaders, reduce the time they have to react, and may foreclose politically sensitive options for de-escalation. Such compressed timelines can only increase the odds of misperception and mistakes, as transparency cannot banish uncertainty or decision makers’ biases.

Transparency’s Double-Edged Sword

One can only hope that into this hot house cooler temperaments prevail and we have leaders that are not driven to mouth off half-baked threats.