Thread by @jheazlewood

Hearing from @timgollins giving a keynote on privacy, identity, accountability and transparency. The private digital archive and tensions over sensitivity #ASA2018

Tim introducing himself by telling his family stories through records and what they leave out or elide. Using this to illustrate the difference between information that is private and information that is sensitive #ASA2018

Talking about EU convention on human rights and the right to be forgotten which is more about identity than privacy because orgs can retain private information but cannot reveal it #ASA2018

What are Archives for? Evidence base for rule of law, hold the executive and the state to account. Holders of semantic capital #ASA2018

What do (government) archives hold? Appraisal in paper world vs appraisal in digital. Probably more information about individuals - ‘case files’, social media - individuals interactions with state #ASA2018

So in this context what is ‘sensitive’? Answer: it depends. Who said/wrote it? When was it said/written? Who was it said to? Who is the subject? What was the context? #ASA2018

Unexpected consequences of this collecting of information about people. Sensitive information may be revealed by the links between pieces of information #ASA2018

Difference between paper based sensitivity reviews (low volume, done by retired experts in field) and digital (high volume, tools not tractable, higher risk, information decays faster) #ASA2018

What tools can we use? Technically assisted review (legal discovery tools). Doesn’t really scale. Columbia University declassification engine - machine learning to work out what wasn’t declassified from what was declassified #ASA2018

Need to work both ends. Supply side (machine assisted review, risk information). Demand side (graduated access moving away from binary open vs closed to allow for more nuance) #ASA2018

Eg Hathi trust allow sandbox access to data which restricts usage outside the environment #ASA2018

Traditional access approach is binary but digital allows for individual privacy while still allowing for access to collective information. E.g. a statistical answer rather than an answer about an individual #ASA2018

Focus should be on sensitivity rather than privacy. Need a calculus similar to what used by national security apparatus but which takes into account all of the answers to the it depends question of earlier #ASA2018

Positing a future where sensitivity reviews, appraisal, modulated access are machine assisted #ASA2018

Interesting factlet, @timgollins says sensitivity reviewers won’t review digital documents but happy to review paper. I wonder if this is bc they are older? #ASA2018

And then I further wonder if this will change over time? #ASA2018

In the interests of open disclosure I am comfortable reviewing online but also feel that tools for that could be improved (ask me! I have ideas!) #ASA2018