Inference is a promise. At the intersection of data and behaviour, inferential moves make ever larger collections of data exciting. A tool of discovery, acts of inference appear to offer an explanatory cause or a future prediction. Sometimes glossed simply as analytics, the nature and use of inference are longstanding mathematical and philosophical debates made newly salient as modellers, simulators, data scientists and mathematicians work to disentangle causation, correlation and reach reasonable conclusions (Franke Program 2021, NAS 2016)
Inference is creepy. Back in 2012, it was clunkily so. The widespread reporting of the US company Target sending coupons for baby items to a secretly pregnant teenager near Minneapolis based only on data from her purchase history led to public discussion of how we are known by companies, and the machines they use (Duhigg 2012). Ten years on, profiling, and prediction through inference in purchase data is less blunt, less seen by consumers.
Inference is promiscuous. Bound up in climate modelling, health data predictions, marketing, news exposure (Thorson et al. 2019) inference is the epistemically perilous work of attributing cause and effect. Its widespread adoption makes it hard to pin down, and few sites arise where those thinking critically about its role and effects might discuss. Nonetheless…
Inference is a meeting point. For growing practices of data extraction, surveillance tools and behavioural analytics. In the workplace, so-called “People Analytics” (Heuvel and Bondarou 2016) take data collected under the guise of ‘efficiency’ and ‘insight’ to correlate worker productivity and background to make inferences about a worker’s future productivity and value to a company (Colclough 2021).
Inference is ethically charged. When data collected on you is used by and for inferences about others, and data on others is used to build a picture of you, acts of inference turn ‘innocent’ data into signposts for things you have not intended or expected to reveal. Through inference, expectations of privacy might be rocked, or undermined entirely. What is done with inferences, verifiable or not, is a decision with ethical consequences.
Inference is in need of governance. An “inference attack” can be used to turn existing data points into an inference about more sensitive information (Gong and Liu 2016). Data can be used to harm. Amassed by large platforms, disparate data have been deployed for predictive ends. Accountabilities for inference are opaque, conclusions are partial, and impacts on lives unclear. Calls for the ‘right to reasonable inference’ in data protection regimes grow (Wachter and Mittelstadt 2019).
Inference is a missing keyword. An inference is not merely a conclusion. It is a process, part both of explicit discourses on what can be known, and a promise of the knowability of the world. I want to make a case for the ethnographic study of inference as discourse and practice. We have long watched the making of evidential practices, explored both scientific and social knowledge in the making (Camic et al 2011). Inference in its social form is more shadowy, more tacit, more absorbed in the background justification to ‘know more’. The leaps, the moves, the connections unseen; the valuations and the reasoning about reasoning matter. The tacit driving promise of inference means it walks largely unnamed through the controversies of our time.
Inference is a knowledge move on the move.
We need to leap with it.
Author: Rachel Douglas-Jones
References
Camic, Charles., Neil Gross and Michèle Lamont. 2011. Social Knowledge in the Making. Chicago: Chicago University Press.
Colclough, Christina J. 2021. It’s not Just About You’. The Why Not Lab. https://www.thewhynotlab.com/post/it-s-not-just-about-you
Duhigg, Charles. 2012. How Companies Learn Your Secrets. The New York Times Magazine.
Franke Program in Science and the Humanities. 2021. Understanding the Nature of Inference: Correlation and Causation, a Multi-disciplinary Exploration. https://inferenceproject.yale.edu/ Yale University, USA.
Gong, Neil Zhenqiang and Bin Liu. 2016. You Are Who You Know and How You Behave: Attribute Inference Attacks via Users’ Social Friends and Behaviours. 25th Usenix Security -Symposiuy, Austin TX, August 10-12.
Heuvel, van den, Sjoerd and Tanya Bondarouk. 2016. The Rise (and Fall) of HR Analytics: A Study into the Future Applications, Valye, Structure and System Support. Article submitted for the 2nd HR Division International Conference (HRIC) on February 20-22, 2016, Australia.
National Academies of Science. 2016. Refining the Concept of Scientific Inference When Working with Big Data: Proceedings of a Workshop – in Brief.
Mayo, Deborah H. and Aris Spanos. (eds). 2010. Error and Inference: Recent Exchanges on Experimental Reasoning, Reliability and the Objectivity and Rationality of Science. Cambridge: Cambridge University Press.
Thorson, Kjerstin., Kelley Cotter, Mel Medeiros and Chankyung Pak. 2019. Algorithmic inference, political interest and exposure to news and politics of Facebook. Information Communication and Society 24(2): 183–200.
Wachter, Sandra and Brent Mittelstadt. 2019. A Right to Reasonable Inferences: Re-Thinking Data Protection Law in the Age of Big Data and AI. Columbia Business Law Review i2.