Nadia: the curious case of the digital missing person
Originally posted by Stephen Easton @ themandarin.com.au
Peter Shergold recently gave public servants a rev-up over the need to actually deliver on innovation instead of just talking about it, but there’s a whole other story behind his prime example, the digital missing person: Nadia.
Nadia is a highly advanced artificially intelligent communication system developed to help National Disability Insurance Scheme participants navigate the new system, but has been shelved indefinitely.
The public administration professor and former public service leader decided it was a perfect example of a cutting-edge project that raised hopes but achieved very little, calling on public servants to vow, “No more Nadias.”
His point was that innovation had to deliver tangible benefits more often — and a lot of people agreed — but his example struck a nerve with one of Nadia’s creators, Marie Johnson.
Shergold suggested it probably demonstrated a failure on many levels, including risk and project management, design, and understanding the needs of users, but in Johnson’s view, this was an “an ignorant and uninformed perspective.”
She explained the “co-design and co-creation effort with people with disability” in great detail last August, in the first submission to an inquiry into NDIS ICT systems (under the name of her business, Centre for Digital Business).
Johnson firmly believes the project is the opposite of a failure, a world-class exemplar of public sector innovation and co-design involving people with disabilities. Her submission suggests Nadia probably could and should be in place now as “an essential component” of the NDIS administration.
In the submission, she argued Nadia was ready to begin training and should never have been put on ice, suggesting this was a poor decision based on risk aversion and a simplistic, incorrect understanding of the project.
As her comments to the committee are long and detailed, covered by parliamentary privilege and on the public record, Johnson told The Mandarin she had nothing to add. So we present parts of her submission below.
“The development of Nadia involved a very deep co-design and co-creation effort based on market and community research; academic research, support and engagement with networks of people with intellectual disability; the development of the operating model within the Agency; and the research and development activities of strategic partners,” Johnson wrote.
The full submission discusses Nadia in her wider context among other examples of the “innovation, new capabilities and co-design” that Johnson believed were vital – not peripheral – to a successful NDIS roll-out. She argues they now appear to have become “lesser priorities” and fears the disability support scheme is facing serious challenges as a result.
She told the committee early inspiration was found in the United Nations Convention on the Rights of Persons with Disability.
“Early in 2015, I led a small but highly capable team that began to investigate what it would take to achieve what the UN Convention describes as ‘augmentative and alternative communication’ and the ability for people to be able to ‘receive and impart information and ideas on an equal basis.’
“I was interested to understand from the World Wide Web Consortium (W3C) what this would mean for the directions of the web. Together with colleagues, I spent time in 2015 with Sir Tim Berners-Lee, W3C Director and inventor of the World Wide Web, and his team discussing the Human Accessible Web initiative.
“The work of the Technology Authority team eventually led to the creation of Nadia, the first digital human for service delivery and co-created with people with disability. Few people outside the team are aware that Nadia’s origins and its very purpose was in the Convention: it did not start as various technologies looking to solve a problem.
“How could it be that people with disabilities, including those with an intellectual disability, could receive and impart information in their own context, and independently?
“Had anyone ever asked or involved them? Had anyone ever acknowledged that the unique insights, skills and experience of people with disability could be imbedded as determinants of design? And that these new design determinants could quickly become mainstream universal design and benefit everyone?
“The Convention is remarkable drafting. It calls out this paternalistic view of treating people with disabilities as ‘objects of charity’ to ‘subjects with rights’ based on informed consent. The realisation of augmentative and alternative communications could only be achieved through the imagination and co-design of people with disability, as demonstrated by the following images.”
She explains how Nadia’s face and the interface she sits within was the product of “deep and iterative co-design” with people with disabilities.
“This human rights-inspired co-design process established the blueprint through which the component technologies – including the AI system and expressive digital human – were brought to life.
“This co-design blueprint encompassed personality, gestures, conversational model, knowledge and market research. University psychology faculty were deeply involved in supporting the co-design with people with intellectual disability, so that the words, expressions and conversational tempo was empathetic and natural.
“Importantly, this supported co-design process ensured that information conveyed through the conversation was understood by people with intellectual disability in their context.
“For the first time, instead of people having to adapt to systems and channels, this was a vision to have systems adapt to people and so go some way to achieving the objectives of the Convention.
“Nadia was an empathetic embodied intelligent digital human, with a deep contextual body of knowledge, able to have a conversation. This was not the case of a simple question and answer chatbot.”
So why was Nadia stood down?
Johnson disagrees with the idea that Nadia was not ready to be introduced, as stated by the current Minister for Digital Transformation Michael Keenan and others.
She went into a huge amount of detail about where the project was up to, the plans for introducing Nadia slowly and training her on the job, the results of very extensive user testing, and the risk mitigation strategies.
Her submission argued there was “no coherency in the argument” about why the project had been shelved and it was “not defensible” based on the facts to claim Nadia was not ready. Her views have not changed. She suggested three elements were missing from the official narrative:
The role of the Department of Human Services:
“Notwithstanding the fact that Nadia was delivered by the NDIA to meet specific NDIA requirements, DHS considered this a whole-of-government capability and the Nadia capability roadmap would be determined by the DHS technology roadmap. From the outset, Nadia was only ever an NDIS capability – co-designed together with people with disability.
“Cate Blanchett agreed for her voice only to be used for the NDIS, not whole-of-government. My advice was that the Nadia capability roadmap could only be determined by the needs of the NDIS participants and this was not a DHS IT role. Obviously, there are indeed whole-of-government learnings and insight from this new capability, including policy. But Nadia was not an IT project, and its progress and roadmap should not have been constrained by DHS IT contracting arrangements.”
The intense fear of government IT failures:
“During this period, there were a number of high profile government IT project failures: Census, robo-debt, ATO. A number of commentators have suggested that the government and the bureaucracy became increasingly risk adverse as a result.
“I believe that this was a contributing factor – but not a reasonable factor – in delaying the introduction of Nadia. The greatest risk in delaying the introduction of Nadia – as forecast in the business case – is being realised: the continuation of “confusopoly”, the impact on operational performance, and the impact of the participant experience and their human rights.”
NDIS political issues:
“Recent reporting in the media and from Senate Estimates, indicates that the NDIA focus is on getting the basics right – i.e. attending to the website, portal and call centre. The problem is, the basics are the problem…
“No amount of additional staff, or work or investment on the website, portal and outsourcing the call centre addresses the fundamental ‘problem’. The fundamental challenge is the circumstances of people with disability and their experience in being supported seamlessly across channels at any time to find information and understand it.
“The PC Report [that proposed the NDIS] – and the Harper Review – called for innovation, new capability and new ways of doing things. The traditional siloed channel by channel approach currently being adopted perpetuates the wicked problem of systemic discrimination so well extolled in the PC Report.”
Johnson’s submission also takes aim at “comments from the bureaucracy” suggesting there was a risk of Nadia going rogue, arguing this is not a concern with the type of AI she is and views along these lines are “a little alarmist, and demonstrate a shallow populist science fiction view of AI”.
“Furthermore, Nadia – as with any AI system – is based on continuous training. Nadia is a new type of system, built on a unique combination of technology and co-designed with participants and stakeholders, underpinned by a cognitive platform. The nature of cognitive systems means that the learning and improvement process is never finished, which means co-design, testing and review for Nadia will be a continuous and ongoing lifecycle.
“This system has been continuously co-designed and tested not only across the technology stack, but through to cognitive dialogue interaction and participant experience, including the experience of people with intellectual disability.
“Functional development of the traineeship Nadia was achieved through rapid cycles of innovation and an agile approach to developing Nadia’s capabilities through extensive co-design, co-creation and testing with the disability community and Nadia technology and academic partners.
“Nadia understands 10,000 question variants, is capable of complex expressions (co-designed with people with intellectual disability), understands spoken and typed questions, speaks with Cate Blanchett’s voice, and can already offer conversations on general NDIS information.
“To achieve this outcome, the NDIA has worked on the boundary of academic research, technology innovation and project delivery capabilities, to infuse co-design and community engagement through all parts of this project. The support and ownership from the community and the DIRG is significant, with members of the DIRG writing blog posts regarding their experience and encouraging this work.”
The submission goes on to say the work on Nadia with IBM, FaceMe and a recent tech start-up called Soul Machines — which was launched as a result of the project — contributed significantly to progress in the field of AI and digital humans.
Johnson told us the following video from February this year was an example of work she had been involved in lately:
The submission noted Soul Machines had showcased its work in Davos while one of FaceMe’s higher-profile projects was a digital clone of the chief economist at UBS, Daniel Kalt.
“And whilst the AI digital human innovation is being deployed by major brands, companies and other governments worldwide, people with disability themselves are told to wait.
“And it is regrettable, that in a desire not to attract too much attention to Nadia, the bureaucracy ordered the withdrawal of the iAward nomination for Nadia, the day the winning iAward for Nadia was to be announced [at the event in 2017].
“So, whilst this innovation was celebrated at Davos, the people with disability who had driven and inspired Nadia’s design were denied recognition and acknowledgement of their remarkable roles in this ground-breaking global achievement.
“This recognition would have provided the people with disability involved in the co-design and more broadly, with potential employment opportunities and growth in the technology sector. And this is the immense social and economic value of the iAwards, as many other people over the 25 years of the iAwards have experienced.
“Being effectively excluded from the recognition and potential employment opportunities through the iAwards is deeply regrettable treatment of the people with disability involved, and the community.”
Johnson argued Nadia’s launch was becoming all the more urgent but warned her introduction was never going to be a simple process, especially after the project being on hold for several years.
“It is not a simple case of flicking the switch on and activating Nadia, as some in the political sphere and in the IT areas of the bureaucracy might believe.
“A complete programme of co-design and the resourcing and skilling-up of an operational model would need to be implemented. Furthermore, both Soul Machines and FaceMe are now offering sophisticated digital humans and therefore vendor selection provides further options to government.
“The Nadia programme should be reinstituted and run by the NDIA as a strategic capability, as envisaged by the business case, independent of DHS.”