Socitm’s Annual Conference brought together the sector’s best, brightest, and most insightful and interesting thinkers – and I was there, too.
One of the many revealing talks featured a panel of luminaries discussing the increasing application of artificial intelligence (AI) in local government services, so I listened in at Leicester City’s King Power stadium on Tuesday.
Maryvonne Hassall, Aylesbury Council’s IT strategy manager, kicked things off by talking about ‘getting rid of the old legacy systems…It’s all about the customer; it’s best to fit residents’ needs to the Kano model: basic expectations, investment, satisfiers, delights and satisfaction’.
She detailed the Vale’s problems with taxi licensing, and how AI has presented a solution. Mary said: ‘Taxi drivers can be demanding, so we have created a streamlined process. On the frontend, a taxi driver can book an MOT test and make payments online.’
The Vale is also using AI in customer service areas – its system monitors questions as they come in, analyses them, and then offers possible answers. A real human customer service agent can then select the best solution suggested by the machine – which learns from the experience to make even better suggestions next time.
Next up was Lancaster City Council’s ICT manager, Chris Riley, who spoke about an onscreen humanoid you can actually interact with. Known as emotive AI, the system uses predictive analytics, shows emotion and predicts what you’re going to say, and then responds.
As well as understanding where a conversation is going, the machine gives out the same info each time – something AI systems haven’t been great at in the past: consistency.
But how near are we to the realisation of this digital assistant? Very near, according to Chris, who presented an image of the thing. ‘The eyes need work – at the moment they look like the eyes of a maniac,’ said Chris.
Considering the costs of this futuristic marvel, Chris added: ‘There are around 400 local authorities, and they’ve each spent on average of £100,000 on their websites. What’s this given us? 400 sites that look remarkably similar.
‘We are not in competition with each other. If we worked together and put just £25,000 in each, that’d be £10,000,000 we could spend on building a good AI system. It’s one interface – we don’t have to develop 400 different machine learning systems. If we do it 400 times it will take a long time to do, but if we all pool the money it will be done much quicker.
‘It has been said to me that I am naïve if I think I can change local government – but local government is naive if it thinks it can’t achieve these things.’
Hilary Simpson, founder of Sleuth the Consultancy Goop, was up next. She noted that the key issue for AI implementation is ethics, and that CIOs should recognise that they have a critical responsibility to expose and help address the potential risks.
‘AI takes us down the dangerous track of profiling individuals based on societal group profiles,’ said Hilary ‘for example, sentencing of black people in Wisconsin automatically leads to longer sentences than for whites’.
She said that fairness and equity must be addressed by public services – and that things are not the same in the private sector.
She noted the importance of taking on board different perspectives and information vs. simple binary automated decisions; for example, in assessing a child potentially at risk.
No technology is neutral, added Hilary, though ‘systems can help. Ultimately, judgement needs the personal touch. The bottom line is the problem of assessing vulnerable people and minority groups.’
But I was back in time to hear Jos Creese, digital analyst, who noted that AI is beginning to take over many aspects of people’s lives. He said: ‘AI does present a number of risks. Mattel have just pulled a babysitting device because of concerns about machines taking over parenting roles. Are we ready for AI to look after things like social care?’
He warned that ‘a lot of jobs will go…it’s about public trust and understanding. If the public sector wants to start using AI as a customer service interface, the public has got to trust us to do it – and I would suggest at the moment they don’t.
‘There are some human aspects to resolve before the technical ones. We are liable to get some things wrong. I guarantee there will be a public outcry at some point when something goes wrong.’
So, a very riveting and illuminating 50 minutes.