Responses for the Blumenstock article

The world is now experiencing a “data revolution”. The popularization of electronic products especially mobile-phones changes people’s life and generates personal data from them every day. Bluemenstock promisingly convinces that data applications based on these vast amount of emerging data can be used as tools to improve humanitarian aid and global responses to crises. For example, through processing personal data from mobile phones especially the number of international calls and imagery from satellites, algorithms can identify people in poverty and match resources to them effectively.

However, Bluemenstock also indicates four noteworthy pitfalls with this powerful tool: unanticipated effects, lack of validation, biased algorithms, and lack of regulation. Firstly, decisions made from big data analysis may not benefit vulnerable people as there are risks of misappropriation. Secondly, the risks and accuracy over time of some new algorithms have not been adequately tested before they are deployed. Thirdly, digital data and the algorithm derived from it are inherently biased, because most disadvantaged people who have no access to internet tend to be the least represented in the dataset. Finally, because of the lack of regulation, companies who only want to maximize their profit may abuse data collected from their users for their own goods.

In order to address these issues, according to Blumenstock, data science need to be considerably more humble which means more attentions should be paid to the people behind the data. New big data analysis methods should be accompanied by conventional data collection methods for validation purpose, customized by taking local context into account, and designed by people who understand the actual problem instead of only the algorithm.

In data science, only possessing a good intent is never enough for making appropriate decisions I believe because sometimes good intent will also lead to unanticipated bad effect. Although big data applications can distribute humanitarian aid in a more focused way, people are able to look for patterns of the algorithms, and make themselves eligible for the aid or profit. As a result, they will take up the resource which should originally be given to people in need, making the aid ineffective.

Moreover, transparency is so important. Unlike convention data collection process, private companies who only want to maximize their profit, generally own and control big data from their users, so they are likely to abuse their data for profit instead of pursuing the interests for most people. For example, when we use our phone number or email address to create an account on some apps, the company will secretly share your information with others. This will lead to so many spams or even crank calls for advertising purposes. Therefore, legislations and regulations to limit their abuse of the power are so important, and increasing transparency will definitely assist such regulatory.

Admittedly, opening every possible data sources with consent is simply the best way to promote data applications globally so researchers can freely use any data needed to develop algorithms. However, hinderances for open data sources will never disappear in the real world. The fear of privacy leakage and data abuse prevents people from sharing their data. In addition, for some political reasons, most countries are also unwillingly to share their data with the world. Hindrances like this can be alleviated by enhancing privacy protection regulatory on the institutions which collect data so people may be more willing to share their data.