2024 UN Civil Society Conference in Support of the Summit of the Future.
Keynote: Raising the Bar on the Ethical Use of Data in Government at the ArabNet Digital Summit 2018
1. Raising the Bar on the
Ethical Use of Data
in Government
Linda Gibbs, Principal
05.01.18
ArabNet Digital Summit – Smart Economy Forum
2. Because who doesn’t love Dilbert?
ArabNet Digital Summit – Smart Economy Forum|
05/01/18
2
3. My Point Of View:
Government has an affirmative obligation to make use
of data
3
• Government that is transparent and accountable
• Relies on evidence to guide action
• Committed to improvement in the public interest
ArabNet Digital Summit – Smart Economy Forum|
05/01/18
5. Government sits on troves of administrative data
– an untapped resource
5
• Data collection is ubiquitous
• Data is highly reliable when used to verify eligibility
• Data is full of knowledge
ArabNet Digital Summit – Smart Economy Forum|
05/01/18
7. But data is rarely shared.
7
• Time and effort to make it available
• Expertise to open it for use
• Cost associated with building the architecture
• Risks associated with data breaches
• Lawyers who say no
• Data “owners” who hoard data for power and control
• Administrators who are averse to sharing facts
• Increasing concern of perpetuating bias
ArabNet Digital Summit – Smart Economy Forum|
05/01/18
9. Data as a Public Good
9
• Culture shift from data hoarding to data sharing
• Data as a public good – clean and open data in the public interest
• Inclusive governance and clear communication
• Privacy and protection as paramount
— De-identification
— System security
• Acknowledging and addressing bias is key
ArabNet Digital Summit – Smart Economy Forum|
05/01/18
11. Data Science has not helped the cause.
11
• Lack of transparency
• Users lack access to their own data
• Privacy has not been protected
• Bias in, bias out
• Ubiquity can terrify
• Lack of trust
ArabNet Digital Summit – Smart Economy Forum|
05/01/18
13. Algorithms with a life of their own.
13
• Bias in the data reflects and amplifies bias in life
• Surrogate indicators not good enough
• Statistical rigor and rigidity of data
• Automated action
• Rush to market, predatory uses
• Surveillance effects magnified
ArabNet Digital Summit – Smart Economy Forum|
05/01/18
15. Ethical Use of Data
15
• Accuracy - Power of sunlight to cleanse data
• User Control - Access and options
• Privacy Protections - Systems security and legal protections
• Transparency - Clearly document methodologies, publish data sets and research questions
• Integrity- Good questions, question causal justification, and careful interpretations of answers
• Diversity among programmers- Lived experience
ArabNet Digital Summit – Smart Economy Forum|
05/01/18
17. Ethical Use of Data - Algorithms
17
• Adequate Math - Statistical rigor on sample size
• Data Relevance - No lazy surrogates
• Careful testing and narrow application
• Constant Data Refresh – to learn and adapt
• Awareness of bias and surveillance effects
• Decision support not decision maker
• Open methodology and available redress
ArabNet Digital Summit – Smart Economy Forum|
05/01/18
19. Do we need a stronger regulatory framework?
19
• Government action where there is market failure
• Where failure occurs, is there strong enough self regulation?
(eg Data for Good Code of Ethics)
• Are public interests such that government action is necessary?
(essential right to privacy and due process)
• Government action – Data Standards, Algorithm Audits, comprehensive
regulatory frameworks
ArabNet Digital Summit – Smart Economy Forum|
05/01/18
Editor's Notes
NEXT
Introductions –
BA
Former DM HHS NYC
History at agency level in child welfare and homelessness, City Hall domain
Experience across those years –
Good intentions enough
Reluctance to be measured by outcomes
Need to move to use of data to guide practice and inform results
The world of data has changed in one generation
Private innovation in use of data has been remarkable
Government data practices have not changed with it
We need pathways to open up the use of data
But it is challenging
There are no clear set of standard practices that governments trust to create that path
It requires expertise, time and money to make it happen
The power sits in the hands of those least likely to want to take the risk – the programs themselves
And we face a public increasingly skeptical about our intentions
So what will it take to move the culture?
define as a public good – a treasured resource that government has a responsibility to unleash
But must be based on integrity of data themselves
And integrity of process which must involve stakeholders
With utmost attention to privacy
And commitment to preventing unintended consequences and undoing bias
Data science has not always been helpful to us
In the hands of private market interests
How’d they know that??? Feeling like every step is watched
Don’t know who has what or how to cut the link
It has often been insensitive or outright biased (like apps to rate “beauty”)
All creating deep lack of trust
And it shows up in scary ways
Increasingly putting algorithms in charge of decision making
Formula not clear – probabilistic associations, sample size
Individuals don’t know how results were achieved and have no power of redress to challenge or correct
Private sector products adopted by government decision makers without fully understanding what drives the methodology
Read from slide
An informed consumer facilitates trust
And data scientists need to be leaders in ensuring ethical use of data – for government use and for private sector use
One informs the other and they are interrelated
If private sector employs poor or abusive practices, everyone suffers
Read form slide
Practitioners need to lead the way – we need a professional code of ethics
And if we don’t we will surely get more governmental regulation
This can be good and necessary
But it can be a rough tool and can stifle innovation
Watching implementation of EU standards which became effective last week
And the fall out from Facebook and Equifax
To see what government needs to do.
Minimally need to take strong, clear, values based action in the private sector regardless, and how the world of data science respects these concerns will be carefully watched