Annie Antón from North Carolina State University’s Department of Computer Science spoke today in the TRUST Seminar on “Incorporating Privacy Values, Policies and Law in Information Systems”. Annie described a series of papers and projects centered around privacy policies and HIPPA privacy.
The first set of work she described involved analyzing privacy policies and a user study involving privacy policies. The set of questions motivating this work includes:
- How do we ensure that system requirements comply with the policy?
- How do we ensure that information handling adheres to policy and system requirements?
To get at the answers to these questions, their team first did a goal-based analysis of a set of privacy policies to pull out teleological goals, strategic goals and tactical goals. This involved a team of three people (a lawyer, a computer scientist and one other disciplinary perspective that escapes me) with a software tool that helps to parse out the goals embedded in the policy. They then used grounded theory to classify the goals into a taxonomy and finally iteratively refined this taxonomy to remove redundancies, etc.
The second study (forthcoming in IEEE Transactions on Engineering Management, 2007) was a user study that involved users reading various treatments of privacy policies and then having their perception and comprehension of the policies measured. They presented users with one of four treatments of privacy policies: the original policy from a website, a list of privacy goals and vulnerabilities, a categorical representation based on their taxonomy (see above) and finally the original policy enhanced such that hovering the cursor over highlighted pieces of the policy exposed the goals of that part of the policy.
Their findings are intriguing and statistically significant to p<0.001.
- Users perceived brand X to best protect their personal information in the highlight case.
- The question “privacy practices are explained thoroughly in the policy I read?” was most agreed with in the original language and highlights cases. (Apparently, the length of a policy leads to a perception of thoroughness.)
- Average comprehension score was highest with categorical, then categorical, then list of goals and vulnerabilities… but, as above, perception was seen to be most secure with the longer policies (this is a paradox).
- Users comprehend the original policies the least.
- There was virtually no correlation in demographic factors, except for people 57+ in age who scored lower in comprehension.
There was more to Dr. Antón’s talk, including systematic analysis of HIPPA, that I’ll have to come back and describe another day in another post.