Challenges and opportunities of digital social research: Access and Anonymity

Transcription

1 Challenges and opportunities of digital social research: Access and Anonymity Dr. Dan Nunan Henley Business School, University of Reading

2 Two narratives for social research: Evolution or revolution? We live in times of great continuity. We can be awfully narcissistic and boring, frozen in the past while making great pronouncements about the present. While we do notice what is changing we don t notice the great many things that are not. - Henry Mintzberg

3 Evolution or revolution? Almost every sector and industry is worrying about the digital revolution an the impact it will have on their work. These predictions are often (predictably) wrong. Given a choice our research shows that 85% of people will skip advertising. This is a devastating problem for advertisers. - MPAA, 1982 Congressional Inquiry (into fastforward button on VHS players). All the things the movie industry feared happened. But they have never been more successful.

4 We are better at predicting what technology looks like than understanding its impact Seated comfortably in the club car of the Twenty-first Century Flyer fast airplane service between London and New York the president of the Ultra National Bank removes a small rubber disk from his vest pocket and places it over his ear. A moment hence, he will receive by radiophone the financial news of the world. Simultaneously, millions of other people all over the globe will receive the message. At designated hours, news of a general character will also be received.

5 New technologies tend to build upon rather than replace existing ones

6 We are better at predicting what technology looks like than understanding it s impact A few common predictions: Radio will replace newspapers. Robots will replace authors. Cinema will replace theatre. Recorded video will replace cinema. The internet will kill TV / books / cinema. And now: Technology will kill survey & qual research

7 What opportunities does the digital evolution create for market research? There is a realisation that the big in big data refers to data collection and storage rather than analysis. Access. What data can we access? Anonymity. What is the role of personal data?

8 Access. Why is it an issue? More than 5,000 online articles contain the phrase: We are collecting more data than ever before but do researchers have more access?

9 What is limiting access? A shift to private online communication and social networks. The information superhighway does not provide the same access rights as the physical highway. The organizations collecting the largest datasets are corporations - not public sector - data that would have once existed in the publicsector sphere is now commercial. Perception that data is the new oil means giving access to data becomes a commercial risk. Promise of commercial API s remains unfulfilled. Data silos - the best data is never shared. In the long run, might big data be small data?

10 Anonymisation. Why is it an issue? - Working with anonymous data is a key identity mark of professional research. - Anonymisation is the key escape from increasing onerous EU regulations on privacy. - If anonymisation fails, trust is lost AOL Data Leak Netflix + IMBD Genomes Project Personal Genome Project Mobile phone aggregation GPS Mining High media coverage but what was the real cost?

11 Anonymization A recent case study highlights the challenges.

12 Anonymization: New York Cab Data Used a cryptographic function that assumes you don t know what the input data looks like. Problem: we know exactly what format a NY taxi driver number looks like. There are only 19 million possible medallion numbers. 2 minutes to compute the medallion numbers. 1 hour to reidentify the entire dataset of 173 million taxi rides.

13 The Anonymization Problem What did we learn? Some celebrities don t tip However, limited practical impact. Relied on multiple open & public data sources. Caused by failure in anonymisation processes. People seeking access to data for illegitimate means have easier and cheaper ways to go about.

14 Two sides to the anonymisation debate. Anonymisation must work. It is a key feature of MR. It is a legal requirement. Without trust in anonymisation we risk a Tragedy of the Data Commons Media stories show how limited the risk of deidentification is. Much of the perceived risk is not related to unethical or possibly illegal activity. vs. Anonymisation can t work. Technically it s not possible to guarantee anonymisation. Claiming anonymisation misleads respondents. Anonymisation is (theoretically) inevitable. Anonymization makes algorithms less useful. How do we reconcile these perspectives?

15 Two sides to the anonymisation debate. Anonymisation of data is a technical & legal solution to a problem that is neither technical nor legal it is social. Anonymisation is about managing risk of disclosure not eliminating it. Managing risk has become a key aspect of modern society - unfortunately, individuals are bad at evaluating risks. They are very bad at evaluating risks that involve complex technology. Trust in social research is determined by how people perceive the risk of submitting data, not the technical aspects of anonymisation If my data gets misused there is risk of harm to what is important to me What is important: Personal safety Family safety Money Work

16 Access & Anonymisation: The Opportunity Source: McKinsey, 2011

17 Access & Anonymisation: The Opportunity Difficulties with access and anonymisation shift the skills need towards those with social research skills (including qual) rather than technicians & scientists.

18 Access & Anonymisation: What next? 1.Perception of the risk of big data is what counts. Communicating about risks of research is critical in building trust. 2.People need to know who they can trust. Respondents need more help than ever differentiating between genuine research & pseudo research. 3.Anonymisation is a promise we might not be able to keep. However, de-identification may be clever, but it might not be legal. 4.If big data turns out to be small data, that s ok.

19 Thank you. twitter.com/dannunan