Machine Learnings Biggest Barriers are still all too Human
“I was about to leave the office a few days ago when an alert popped up on my phone warning me of potential delays on my journey home. Such alerts have now become commonplace and have lost their initial ‘wow’ factor, but it’s a useful illustration of how far we have come in the past couple of decades. What would have seemed like science fiction at the start of the century is now a day-to-day reality, provided you are prepared to share some of your data.
Continuing advances in machine learning technology should allow us to reap even greater benefits in the years ahead – with highly personalised, real-time transport data, alerts and services beamed directly to our mobile devices, allowing us to minimise disruptions and improve our travel experiences no matter how complex our journey and no matter how many different forms of transport we are using.
By combining details about our own travel needs and preferences, live information from a rich variety of data sets, and accurate system-wide predictions based on machine learning analysis of that live data and historical patterns, we can expect to receive updates and advice that go far beyond telling us that the M4 is closed between junctions seven and eight.
Already we are seeing journey planners that can infer journeys from the calendars in our smartphones. As these mature, we will see them offering us more personalised options for getting from Wednesday’s meeting in Leeds to Thursday’s conference in Cardiff and, if we wish, taking care of the overnight hotel booking at the same time.
Our regular journey preferences will also be well known to our virtual journey planner. It will know if we usually like to cycle to work but will also warn us when heavy rain is forecast and offer to book us a taxi or remind us where we can get a bus from.
The planner may even be aware of differences between business and leisure travel – and offer to pay for flights using our frequent flyer miles rather than the usual credit card when it works out that the trip to Florida in July is a family vacation.
But again, this will rely on people and organisations as much as on technology. Computers will continue to get faster, smarter and more powerful – but if we are unwilling or excessively wary about feeding them with the data they need to make their calculations, the outputs we get back will be little better than the relatively crude approximations that still inform the current generation of transport information platforms.
As a not-for-profit, neutral player in the burgeoning Intelligent Mobility sector, the Transport Systems Catapult has long championed the benefits of sharing data, while of course preserving individual privacies and still allowing data-owners (whether people or organisations) to decide how that data is used, and by whom.
Ultimately, people will only be willing to share their data if they can see that doing so brings them personal benefits. Indeed, while Intelligent Mobility relies heavily on technology to achieve its aims, the aims themselves are all about putting people (in our case, transport customers) at the centre of the latest transport innovations.
Reasons to be share-ful
While the end goal is to have travel apps that are free and easy to navigate, I am by no means suggesting that the creation of multimodal, highly personalised journey planners will be a simple task - particularly since the software will need to draw on information from a wide variety of data providers and be able to adapt to rapidly changing circumstances. But unless they can gain access to the raw data in the first place, programmers will find it difficult to even make a start on addressing these remaining complexities.
Fortunately, most of the transport providers who we work with at the Transport Systems Catapult are already switching on to the benefits of shared data – even if many still have understandable questions about how they can best share it while still protecting their commercial interests and potential revenue sources.
It is something of a chicken and egg situation, with many of the best business cases for sharing data with other organisations (including, sometimes, commercial rivals) often only evident once the sharing process has begun.
Generally though, it should be clear that the benefits of sharing data will outweigh the commercial risks which will be faced by organisations that resist making their data available to partners and/or third party developers. Fuelled by continuing smartphone penetration and further growth in mobile connectivity, new technology platforms will be able to offer far smarter solutions to customers who will also increasingly see the value in sharing certain anonymised elements of their personal data.
This last point was underlined by the highly detailed ‘Traveller Needs and UK Capability Study’ which was carried out by the Transport Systems Catapult in 2015. Combining a survey of more than 10,000 travellers with the expert opinion of industry leaders from over 70 cross-sector organisations, the report identified three core areas where travellers are particularly keen to see improvements:
- Enabling lifestyle fit – acknowledging that there are different types of travellers and supporting them with mobility options that meet their individual demands;
- Enhancing end-to-end journeys – actively engaging travellers in planning their journeys, with flexible and convenient options that meet the context of their travel, taking into account all available forms of transport;
- Removing pain points – including an easing of parking challenges and the enabling of smoother drives and enhanced multi-modal journeys.
All three of these areas were identified by regular travellers as key concerns that they wanted to see addressed. This would strongly suggest that they would be prepared to share the relevant data in return for the increasingly sophisticated solutions that the next generation of apps are expected to provide. It is this combination of new technological capabilities (fuelled by machine learning and other technological advances) and new data (willingly provided by customers and transport operators) that will allow for truly revolutionary transport services, while also creating the business demand to support their development.
As my earlier example with the single-mode traffic alert demonstrates, we are already happy to receive up-to-date information that can improve our journeys – or at least explain the reason behind a delay – even if the alerts are still relatively basic. So the demand for ever more sophisticated alerts is surely there, even if the supply is lagging behind.
In a similar vein, it was striking to attend the recent Billion Journey Project demonstration days organised by the Go-Ahead group and see how all ten of the start-ups being showcased were reliant on data to power their services and solutions, with many also including machine learning and/or artificial intelligence as part of their product design. Wluper and Enterprise Bot, for example, are both building automated ‘chat bots’ to answer traveller queries in a natural conversational style. Citi Logik meanwhile presented an information service that can predict the likelihood of seats being available on the next train, while FAIRTIQ demonstrated a smartphone-based app that allows passengers to buy a virtual pass that is valid across a variety of transport modes.
In short, we can already see many of the ideas and the technologies that will drive the transport apps of the future. But for these apps to work at their very best, we need to keep them supplied with a steady stream of useful and accurate data.
What’s holding us back?
I have already touched on some of the obstacles to the wider sharing of transport data – the most notable being the privacy concerns of individuals and the commercial concerns of organisations – but I am happy to say that progress is also being made in this area.
The Transport Systems Catapult has long advocated the creation of a market place for transport data and in the last couple of years we have started to see examples of this concept emerging, for example in the automotive sector with organisations such as Caruso and Otonomo, or, the recently announced £8m Meridian 2 project to create a UK facility for the exchange of connected and autonomous vehicle data.
The success of ventures like these will help lay the foundations for the next generation of applications to support and improve travellers’ journeys – while hopefully also demonstrating the customer and commercial benefits of making data available, either freely or at a competitive price.
I am fairly optimistic that in ten years’ time we will look back with nostalgia at the basic text alerts that told us the motorway was congested or that the train was delayed but didn’t tell us what we could do about it. If, however, we are still stuck with single-mode, rudimentary updates, I suspect the blame will lie not in a lack of data but rather our own failure to set the data free.”
In the next article, Luisa Moisio, RSSB’s Research Programme Director, will evoke the do’s and don’ts from a business strategy perspective, that rail decision makers will need to consider on their Machine learning deployment journey.