Algorithmic bias or fairness: the importance of the economic context – Catherine Tucker

MIT Sloan Distinguished Professor of Management and Professor of Marketing Catherine Tucker

From the Shorenstein Center

As a society, we have shifted from a world where policy fears are focused on the ubiquity of digital data, to one where those concerns now center on the potential harm caused by the automated processing of this data. Given this, I find it useful as an economist to investigate what leads algorithms to reach apparently biased results—and whether there are causes grounded in economics.

Excellent work from the discipline of computer science has already documented apparent bias in the algorithmic delivery of internet advertising [1]. Recent research of mine built on this finding by running a field test on Facebook (and replicated on Google and Twitter), which revealed that an ad promoting careers in science, technology, engineering, and math (STEM) was shown to between 20 and 40 percent more men than women across different age groups [2]. This test accounted for users from 190 different countries, with the ad displayed to at least 5,000 eyeballs in each country. In every case, the ad was specified as gender-neutral in terms of who it should be shown to.

When my team and I investigated why it was shown to far more men than women, we found that it is not because men use these internet sites more than women. Nor is it because women fail to show interest or click on these types of ads—thereby prompting the algorithm to respond to a perceived lack of interest. (In fact, our results showed that if women do see a STEM career ad, they are more likely than men to click on it.) Nor does it seem to echo any cultural bias against women in the workplace. The extent of female equality in each of the countries as measured by the World Bank was found to be empirically irrelevant for predicting this bias.

Instead, we discovered that the reason this variety of ad is shown to more men than women is because other types of advertisers actually seem to value the opportunity to get their ads in front of female (rather than male) eyeballs—and they’ll spend more to do it. Some advertisers’ willingness to pay more to show ads to women means that an ad which doesn’t specify a gender target is shown to fewer women than men. In essence, the algorithm in this case was designed to minimize costs and maximize exposure, so it shows the ad in question to fewer expensive women than what amounts to a greater number of relatively cheaper men.

Read More »

Increasing click-through rates with ad morphing — Glen Urban and John Hauser

MIT Sloan Professor Glen Urban

MIT Sloan Professor Glen Urban

MIT Sloan Professor John Hauser

MIT Sloan Professor John Hauser

From Fortune China

Everyone is trying to make their banner ads and new media more effective. In the banner area, 90% of the effort is spent on targeting. If you click on a link, you’ll get a particular ad. A whole industry has emerged focused on collecting click stream data and making recommendations.

But that is only half the picture. Equally important is the question of how you should talk to consumers once they are targeted. This is what ad morphing is all about.

For example, a car company may target a consumer whose click history indicates he is interested in buying a car. However, instead of just randomly sending him car ads, it can track the consumer’s online behavior to determine his preferred communication style. We also call this his cognitive or thinking style. Does the consumer want a picture of the car at a NASCAR race? Or would the consumer prefer to look at the technical aspects of the engine? Or does the consumer want a fashion shot of a driver pulling up to a country club? What will the consumer best respond to?

This is a multi-arm bandit problem because it’s like a slot machine with many arms. The advertiser needs to choose the ideal lever to pull to match the ad to the consumer’s thinking style. However, it’s more difficult with ads because there is uncertainty as to the consumer’s thinking style.

Our algorithm addresses this issue by monitoring click stream data to determine how a consumer makes decisions on the web. After enough information is gathered, the algorithm determines the consumer’s likely thinking styles and matches the optimal ad to our estimates of thinking styles – all in real time.

Partnering with companies like General Motors to test our algorithm, we found that morphing has tremendous potential to increase banner ads’ productivity. Companies work hard just to get a 1-2% improvement in click-throughs, but we found that morphing ads based on thinking styles can improve that rate up to 83%. We also found that morphing can lead to 30% better brand recognition. These are very significant effects.

While our algorithms (see our paper for the algorithms) can be implemented by any good programmer skilled in the art, morphing can challenge the budget. To use this tool, companies have to design more ads – ads that appeal to each of the various thinking styles of customers. There also may be cross-organizational issues, as the people who create those ads must coordinate with the analysts doing the targeting.

However, this is the only algorithm that we’re aware of that integrates thinking styles and morphing in real time. It’s very cutting edge, but it can help move the market to the next wave of action in banner advertising.

Also see the post in Chinese at Fortune China.

Glen Urban is the David Austin Professor in Management, Emeritus, Professor of Marketing, Emeritus, Dean Emeritus, and Chair of the MIT Center for Digital Business at the MIT Sloan School of Management. 

John Hauser is the Kirin Professor of Marketing and a Professor of Marketing at the MIT Sloan School of Management. 

MIT Sloan grad on the “sharing economy,” the next big trend in social commerce

Jaime Contreras MBA ’11 and Tal Snir MBA ’11

My first entrepreneurial venture in the so-called “sharing economy” happened accidentally. It was 2009, and I had just moved to Cambridge, MA to start at MIT Sloan. I brought my car with me because I was sure I would need it. As it turned out, the only time I used it was to go grocery shopping; mostly it sat idle. Word got around that I had a car, and I soon found myself fielding requests to borrow it from friends. One suggested I charge for the privilege.

Read More »

Duncan Simester on customer response: Why guess and get it wrong when you could do a little experimenting and get it right?

MIT Sloan Prof. Duncan Simester

The business pages are filled with examples of companies that have taken big hits to their brands because they’ve made marketing decisions that ran afoul of customer expectations. Take Netflix, and its aborted scheme to divide its streaming and DVD video offerings. Netflix could have avoided its embarrassing reversal if it had experimented on this decision before publically announcing the change.

Read More »

How to improve products? Survey consumers with "active machine learning"

MIT Sloan Prof. John Hauser

When you buy a house, it would be irrational to search every possible house on the market. Instead, you narrow down your choices based on things like price, location, and number of bedrooms. The same thing happens when you buy a car. You might only look at sporty coupes or hybrid vehicles. Everyone has their own individual methods – or heuristic decision rules — for screening products, usually based on the item’s key features.

This presents a significant question for companies:  How do you determine what these decision rules are? Managers are increasingly interested in this topic as companies focus product development and marketing efforts to get consumers to consider their products or prevent them from rejecting the products without evaluation. If they better understood consumers’ heuristic decision rules, they could use this information in the design and marketing of new products.

Read More »