Funding opportunity: Nesta has launched a £100k fund for ‘AI for Social Good‘ applications.
This article proposes a potential use case of ‘No Wrong Door Chatbots’, a way of harnessing AI for Social Good to better help homelessness support services.
The Cloud can make powerful AI capabilities more accessible – For example Microsoft offers their Azure AI suite, Google ‘Explainable AI‘ and AWS a Machine Learning Cloud service, making it easier to harness capabilities like AI predictions.
With Scotland defining the exciting ambition to develop an AI strategy for the country, the key question is like for any new technology adoption, what quantifiable business benefits will it bring and more importantly what transformational social good can it achieve?
As part of a national strategy, how can we develop a common model for transformation that could be applied across multiple scenarios?
AI for Government – Automating efficiency
First we can consider the building block of Government’s adopting the technology for improved process efficiency, highlighting that intelligent software can take many forms including simple automation, through ‘RPA’ – Robotic Process Automation.
As Phil Vincenzes describes writing for GovLoop the key dynamics to look for when considering where RPA might be applied are manual processes where they are highly repetitive, time sensitive and expensively error prone, achieving a quick, quantifiable ROI.
In Scotland an example is North Lanarkshire Council. Our case study explains how they are looking to apply the technology for this type of process improvement:
“The back office is fully digitised. Robotic Process Automation (RPA) is a common feature at the Council, taking over repetitive rules-based tasks e.g. invoice payments.”
as part of a general adoption of AI capabilities across the Council:
” artificial intelligence (AI) can already be used to respond to customer queries through chatbot features or even virtual assistants. By 2029, it is predicted AI will be at about the same level of intelligence as adult humans. This technology and others unlock possibilities for delivering services in radically different ways – if we invest in the necessary infrastructure.”
AI for Social Good – Harnessing Intelligent Systems
Of course any government efficiencies are better for the public but it would be a stretch to quantify this as social good.
To begin mapping out how this might be defined we can consider the broader ecosystem that local government organizations like North Lanarkshire participate in and do so within a context of Scotland’s ambition of achieving Integrated Health and Social Care.
This requires local councils like North Lanarkshire to work as part of a local Integration Authority, the intersection of local government with Healthcare, so that citizens are treated through a holistic model that blends their previously isolated departmental services into a single framework.
This requires them to work collaboratively with local ‘Third Sector’ organizations, and it is this area where many challenges arise. How they are all unified into a single delivery unit is the overall goal, and where technology like AI can be applied in a powerful way.
While we tend to think of AI being applied to a single application scenario, it’s true potential will be unleashed when it serves to increase the ‘systems intelligence’ of these types of multi-organization collaborations, making their touch point interfaces simpler and their combined services smarter and better integrated.
No Wrong Door Chatbots
A simple example of this effect is the idea of “No Wrong Door Chatbots”.
No Wrong Door is a simple but very powerful concept that illustrates the key ideals of this article. As the name suggests it means no matter which organization a citizen makes contact with, it will provide the same consistent experience. Ie. there is a common systems intelligence uniform across them all.
However the practical challenge is that every one will have experienced the frustration of being ‘passed from pillar to post’, common for highly departmentalized organizations like government, with staff ill-equipped with the knowledge required for the ‘sign posting’ and referrals to handle cross-agency case work.
This is especially challenging for complex social needs like homelessness. The Scottish Government’s Homelessness and Rough Sleeping Action Group defines:
“Create a ‘no wrong door’ approach where front-line public services act on behalf of or in concert with Housing Options, and the flexibility to delegate to commissioned services and voluntary sector where appropriate.”
Chatbots offer an example of how to tackle this challenge and manifest the type of networked intelligence described here. It’s a great customer service tool, offering embedded, instantaneous responsiveness on web sites.
However “out the box” they aren’t the answer – Similarly many people will have experienced how they can just be ‘dumb’ agents, pre-programmed with only a small number of response options, and in effect really only acting as a ‘press 1 for this, press 2 for that’ menu navigation interface like you experience from call centres.
Hence the point about AI being utilized more for engendering a systemic intelligence not just an isolated app. What it can achieve and how it is ‘programmed’, what body of knowledge it works from, is the essential dynamic.
This is exemplified through this story of an enterprising teenager who set out to apply chatbot technology to the homeless challenge, describing it’s fairly limited functionality:
“Most commonly, this means drafting a legal letter that the person using the chatbot can send to their local council to apply for emergency housing.”
“Browder says he used data gathered via the Freedom of Information Act to figure out trends in why public housing applications are approved or denied, as well as from a team of volunteer lawyers. This is so the bot can create the best possible application for each individual.”
In the case of the Shelter Scotland, and these chatbots, the question to explore is what body of knowledge are they based upon? How does that knowledge accurately reflect all of the ecosystem of the organizations who provide services relevant to that need?
The more these interface points are based upon data that reflects this networked and interconnected industry, the ‘smarter’ they will be in terms of being able to respond with effective responses and help.
The knowledge that will be helpful to citizens needs is distributed across multiple organizations, their processes and subject matter experts. If an organization approaches chatbots as interfaces to only their services they will simply automate the silo effect.
If instead they collaborate to create a pooled domain knowledge, a networked intelligence, and program chatbots with that bank of information, they create an intelligent touch point that can literally function as a No Wrong Door capability.
A critical point is that then this capability can be shared across all of them as a common function. Each organization can integrate it into their web site so that each offers this consistent engagement method, each offering the same No Wrong Door networked intelligence.
It could be available on this page and every other touch point for the homeless; it would also make for an ideal mobile app for outreach workers too, who could make use of it on behalf of those they are helping.
Clearly the potential for this technology is immense. This is just one simple example to illustrate some ideas for methodology for how it might be best applied. Google describe over 2,600 potential use case scenarios, with organizations like AI for Social Good Foundation and the Turing Institute offering deep programs to support scaled implementation.
For Scotland’s AI strategy the essential dynamic is to ensure the program is aligned to the real needs of the nation’s populace. Programs like Integrated Health and Social Care are highly fertile grounds for developing solutions of real value, and the country is experiencing a rising homelessness crisis.
The technology will truly realize it’s full potential and promise when it can be applied to achieve meaningful impact in these areas, and this one suggested way in which the strategy could be directed towards that goal.