Should artificial intelligence and rail ticketing be coupled?

From a technical perspective, artificial intelligence is a theme that runs through many of my blog articles. Not surprising given the fact AI is such a big talking point, especially in the world of IT.

But a non-technical subject that’s cropped up in one of my recent articles is the railways. And while that topic is more unexpected for a solution architect’s musings, it is the 200th anniversary of the invention of the modern railway this year, so definitely worthy of reference.

There are multiple areas where artificial intelligence and railways could combine, especially in the face of disruption: timetabling, track capacity, station availability, staff rostering, etc. These all seem ideal candidates for the kind of multi-input, multi-output decision-making to which AI is particularly suited.

(As an aside, AI is already being used in the rail industry, such as Network Rail using the technology to proactively find and repair faults along the railway.)

Could AI meet its match in the UK rail ticketing system?

Let’s look at the complications of the UK rail ticketing system and consider whether this is where artificial intelligence could meet its match.

What feels like it should be a fairly straightforward process – where prices are based on how far, how soon, how nicely – turns into a myriad of overlapping, outdated possibilities. These are initially based on all those criteria, but then many more factors are added, the most impenetrable of which is the number of journeys. For example, two short ones can cost less than one long one over the same distance.

Add in the variations caused by having to use multiple operators for a single journey, and prices almost feel personalised, whereby two travellers often end paying a different price to travel between the same start and finish points. And the revenue share models are equally difficult.

To paraphrase the late writer Raymond Chandler, perhaps this is the biggest waste of human (or artificial) intelligence you can find outside an advertising agency.

Complex doesn’t have to mean complicated

The ‘network’ nature of a railway means its operation is inherently complex. Everything is connected to everything else (often literally), and the number of variables means a problem in one area inevitably has unexpected consequences in others.

However, from a user perspective, the ideal scenario could require going yet one stage further. Imagine being able to drive to a local park-and-ride, park, catch a tram to a mainline station, travel to a long-distance destination, and then complete the journey by hire bike – all based on a single payment. Ideal from a customer perspective but a potential nightmare in terms of organisation.

And while providing this level of variation may indeed be complex, it doesn’t necessarily have to be complicated. The variations introduced by the number of restrictions means it’s almost impossible to predict in advance what could happen, hence the complexity. But the complication is introduced by further variations that result from adding well-known-but-intricate rules. The complexity is inevitable; the complication is optional.

That means there’s actually a fairly straightforward way to avoid complication, and that is: minimise the number of arbitrary rules (railway ticketing, take note!).

Why good information is key

But avoiding complexity takes more work. Or more accurately, more patience. Although the combinations of a large number of parameters in a network/system mean a lot of things could happen, in practice, not all, and often not even many, of those combinations do happen.

So, while it may not be possible to make a prediction about what will happen, it is possible to observe what has happened. In effect, the system’s behaviour emerges over time.

To identify this ‘emergent’ behaviour requires good information. This could be seen as needing to define a data model for the network/system but the nature of complexity means that’s unlikely to be possible (at least accurately).

Instead, recording interactions that capture the results of activities and processes performed should be the primary technique to use when trying to identify emergent behaviour. Once those interactions are clear, the data – and associated detailed functionality models – will come from them.

An interaction-oriented approach

Working from a definition of the interactions recorded/controlled makes it easier to move from observed behavioural patterns to business process descriptions. And the objects/data – as well as the functions needed to manipulate them – are then seen as two subsidiary parts of that higher-interaction concept. In turn, this interaction-oriented approach allows a solution to develop more naturally than by starting with assumed concepts that immediately require detail/precision.

Quantiv’s NumberCloud service and NumberWorks method are specifically designed to assist this approach. NumberCloud supports the collection of metrics without the need for predefinition of the data to be collected, while the NumberWorks method allows the behavioural patterns that emerge to be described in a consistent format.

So, even if complexity in railway operations is inevitable, that doesn’t mean the network has to be unmanageable.

But the complications are entirely the making of the organisations involved.

Talk to our team

To find out how NumberCloud and NumberWorks could support your organisation, contact us today on 0161 927 4000 or email: info@quantiv.com