Facial recognition technology is attracting a lot of headlines at the moment, and for all the wrong reasons. What’s all the fuss about? Should we be eagerly looking for ways to deploy it or steering well clear?
Firstly, why, to some people at least, is it so troubling?
Fundamentally, because it feels exceptionally personal, and the perception that it can take place anywhere at any time without you knowing feels disturbingly intrusive. We are used to unseen activity in the digital world, but in the physical world it is more problematic, and questions of consent more emotive. Furthermore, any data can potentially be abused, sold or hacked, and biometric data such as your face is particularly worrying in this respect: unlike your password, you can’t change your face. Finally, I don’t think we can ignore the fact that the technology has strong associations with policing and repressive states, and those perceptions can taint more benevolent usage.
But … it is potentially very useful. It's a completely friction-free process that provides a strong form of identification and other useful information without any user interaction, and therefore has the potential to be very convenient.
As with any new technology, there is hype, there are gimmicks, there are risks, and there is exaggeration of all these things. The crucial thing is to think about what it is genuinely useful for in the context of a business.
Let’s set aside the more obvious crime prevention functions that attract most of the headlines and look at some other possible use cases. One important point to note when thinking about these is the distinction between true facial recognition, which implies the face is being compared to a stored database of known individuals, and facial scanning, which need not. The latter could simply detect general characteristics such as age and gender, or perhaps emotional cues, reducing data protection concerns.
- Access control - Speeding up entry point bottlenecks such as stadium turnstiles, theatre doors, boarding aircraft, etc.
- Transaction authentication - Queue busting and quick checkout at controlled locations such as food ordering kiosks.
- Customer service - Recognising known customers so as to provide better service, such as greeting them for a pre-booked appointment or goods pick up (rather like ANPR automatically recognises cars for a toll road or ferry).
- Age verification - For example to avoid the need for a manual approval to buy restricted goods – a great example of a use case that doesn’t need the individual to be identified and solves a clear point of customer friction.
- Car safety - For example monitoring driver wakefulness, a reminder that facial technology also has private uses as well as the more problematic public ones.
- Health conditions - Tools for the individual and/or medical professionals that detect changes in appearance relating to illnesses and predispositions.
- Tailored marketing - Recognising customers either individually or by demographic group to deliver more relevant marketing, and potentially connect to loyalty schemes.
- Behaviour analysis - More general recognition of customer and employee behaviour patterns, potentially including emotional reactions, when people are in e.g. a retail space, to build up an aggregate picture of behaviour, supporting decisions around product placement etc. and/or to optimise operational factors such as queues and overcrowding.
As with any new technology, the crucial thing is to look past the gimmicks and freakouts and uncover what it is actually useful for. The starting point for that should be thinking about what benefits your customers, looking at their needs, attitudes and touchpoints. Are there ways such as those listed above – purely intended as suggestions to provoke thought – in which facial recognition or scanning can give them a better, faster or more convenient experience?
Once you have a good potential use case, a few general considerations emerge in thinking about mitigating possible risks:
- Communicable benefits - Are the benefits to the people being scanned clear, immediate and easily communicable?
- Informed consent - Genuine consent is crucial to avoid the risk of a backlash if people become aware of what has been happening too late. Is your intention totally transparent, with people really understanding what they are agreeing to?
- Trust - Think honestly about the level of trust you have as a brand or organisation. Will people believe your good intentions?
- The human factor (1) - Does your use case square with the natural rules of human interaction? Feel your way through the experience in the shoes of different people and personalities. Will it be a moment of convenience or delight or will people be thinking "oh … they're watching me … how do they know that?". Could it set up problems for your staff?
- The human factor (2) - A subtler angle on this point. Is your technology dehumanising rather than genuinely improving interactions between customers and/or employees?
- Discrimination - Finally, beware of the dangers of accientally discrimating on e.g. race / sex grounds because of the (real or perceived) biases of machine learning algorithms.
One of the dangers with issues such as consent and creepiness is that individual reactions are so varied. Your opinions may be way out of line. If in doubt, run an experiment.
To summarise, yes, there absolutely are scenarios where facial recognition technologies can be implemented and bring value. But it all depends on looking hard at your customer experience and finding the genuine benefit that will be compelling for your customers. Overall, do you have a nugget of customer value that can take the conversation beyond crime and creepiness to something that will make people smile for those cameras?
Author: Tim Johnson