Summary and Moving Forward
We’re now operating in a world where automated algorithms make impactful decisions that can and do amplify the power of business and government. I’ve argued in this paper that we need to do better in deciphering the contours of that power. As algorithms come to regulate society and perhaps even implement law directly,47 we should proceed with caution and think carefully about how we choose to regulate them back.48 Journalists might productively offer themselves as a check and balance on algorithmic power while the legislative regulation of algorithms takes shape over a longer time horizon.
In this paper I’ve offered a basis for understanding algorithmic power in terms of the types of decisions algorithms make in prioritizing, classifying, associating, and filtering information. Understanding those wellsprings of algorithmic power suggests a number of diagnostic questions that further inform a more critical stance toward algorithms. Given the challenges to effectively employing transparency for algorithms, namely trade secrets, the consequences of manipulation, and the cognitive overhead of complexity, I propose that journalists might effectively engage with algorithms through a process of reverse engineering. By understanding the input-output relationships of an algorithm we can start to develop stories about how that algorithm operates.
Sure, there are challenges here too: legal, ethical, and technical, but reverse engineering is another tactic for the tool belt—a technique that has already shown it can be useful at times. Next time you hear about software or an algorithm being used to help make a decision, you might get critical and start asking questions about how that software could be affecting outcomes. Try to FOIA it, try to understand whether you can reverse engineer it, and when you’re finished, write up your method for how you got there. By method-sharing we’ll expand our ability to replicate these types of stories, and, over time, perhaps even develop enough expertise to suggest standards for algorithmic transparency that acknowledge business concerns while still surfacing useful information for the public.