Artificial Intelligence in Service Design
AI in the spotlight
Technology is more than ever impacting our daily lives. Services are fundamentally shifting from mere human-to-human contact towards a mix of technological and human interaction. Designers should already take this evolution into account when creating future experiences. In this think-piece we will spotlight artificial intelligence, its accompanying design hurdles and how to overcome them.
Level of Intelligence
Many different perceptions on artificial intelligence exist, but they all relate to machines or technologies being able to make decisions and think to some extent like human beings. We speak of the term Narrow Artificial Intelligence (NAI) when AI is capable of doing a single task as well as a human being1. Think of the AI Program that beat the best AlphaGo player in the world or think of real-time weather forecasts and personalised purchase recommendations we have today.
This intelligence could eventually advance into General Artificial Intelligence (GAI) only when it is able to reason completely like a human being would. We have not yet been able to reach this level of intelligence and it is still highly debatable if and when we would be able to reach this. Once achieved, this could further evolve into Super Artificial Intelligence (SAI) where AI becomes much smarter than the brightest human brains in almost every field. This is often portrayed in movies in a dystopian future where technology takes over the world. Nobody really knows how fast intelligence will evolve, whether it will be beneficial or disastrous or even if we will ever get this far. Some theorists claim that we are getting closer to GAI and that once we have reached this state, we will exponentially and almost immediately reach SAI as a consequence.
Augmented Human Intelligence
As artificial intelligence and other technologies evolve, the boundaries between technology and humans dissolve and the line of interaction becomes less distinct. Google’s assistant can already understand the nuances of real-time conversations, respond naturally and even ring up a salon or a restaurant to make an appointment for you. China recently introduced the world’s first artificial intelligence news anchor at the World Internet Conference2. It learns from live videos and is able to continuously work without having to take any breaks.
We connect more than ever with technology, but service designers often still commit to a viewpoint where back-end infrastructure and technology merely support the service. This mentality can for instance be found in service blue prints where systems and technology are placed at the bottom within support processes.
Technology will keep evolving and the need to collaborate with it rather than to only see it as support will grow with it. Assistive AI or Augmented Human Intelligence looks at AI as an enabler of our intelligence, rather than a separate competitor to the human brain. To fully take advantage of our enhanced intelligence, we should bear in mind potential hurdles and how to overcome them in future service design.
Hurdles to overcome
In Machine Learning (NAI) computers have the ability to learn by themselves without explicitly being programmed. In a potential future, a healthcare program could for example diagnose patients with a certain disease based on historic data it has gathered from a large set of patients. The system will automatically generate complex algorithms and predict or calculate certain outcomes based on their initially given input and expected outcomes. However, they seldom provide any logic in how they have reached these end-states. This lack of transparency makes it difficult to fully trust these black-box models3.
Moreover, the data used to train these models often has an intrinsic bias from the imperfect real world. We risk to reproduce prejudice if we don’t have a critical look at these algorithms, as this could lead to lumping individuals into wider categories with potentially unfair consequences3. The COMPAS algorithm for example calculated the potential risk of reoffending violent crimes based on the US states’ judicial systems. Not only did it turn out to have a very low accuracy rate, it also included unfair biases such as scoring black defendants at a higher risk than white defendants. Similarly, Amazon had to scrap its secret AI recruiting tool as it showed bias towards women4.
Artificial Intelligence will further reduce human error in the long run, which would be beneficial for the shown case on diagnosing patients. It could also greatly reduce personal biases when used in the context of research. However, sometimes this human error is exactly what endows an experience with authenticity3. In services it is often this human error that we relate to authentic experiences5. When reducing human error to the maximum, it will be a challenge to create genuine and unique experiences that truly delight the end user.
Ethics related to artificial intelligence are already highly discussed and will become even more relevant when approaching higher levels of intelligence. A self-driving car could come to a point where it has to make the impossible decision between crashing its passengers or driving into the human being crossing the street in its path. Artificial intelligence is already approaching high-fidelity realism by representing it in simulated voice and video technology. Gustav Borgefalk indicates that already people’s interactions with virtual assistant are often considered abusive5. The case of sex robots is perhaps the most illustrative example of how humanoid robots may be abused. It is important to understand whether abuse towards artificial Intelligence reinforces this behaviour towards other human beings.
Reimagining Service Design
Artificial intelligence should no longer be seen only as support, but as a potent enabler, symmetrical participant6 or even as an equal new stakeholder7. The need for two-way communication between technology and humans arises.
Established techniques such as service blueprinting or stakeholder mapping and other sworn by frameworks will have to be reimagined to align with the suggested two-way view. As a service designer it is essential to stay informed about these emerging and enabling technologies. While being acquainted with recent technology developments, it will become more apparent when to include technological profiles in co-creation sessions.
Superb interfaces will also need to be put in place to enable efficient human-to-machine communication. People will have the need to be able to comprehensively communicate to systems when wanting to feed instructions and information. Waze, the world’s largest community-based traffic and navigation application, uses input from drivers to optimise its real-time traffic information. On the other hand, first-rate translation of underlying artificial reasoning is required to enable full transparency and understanding.
Lastly, regarding morality in relation to this two-way communication, robo-ethics8 are already considering how artificially intelligent beings may be used to harm and benefit humans. Science fiction author Isaac Asimov9 already introduced the three laws of robotics in 1942. It is likely that in the future, we would have modern legal and moral obligations towards our machines, similar to human rights or animal rights.
When designing new or improved services, bear in mind potential new technologies which could function as an enabler for your service instead of mere support. When this is the case, involve the right profiles in co-creation sessions and adjust frameworks and techniques accordingly.
Acknowledge potential hurdles that could come with artificial intelligence and prepare for a transparent two-way communication with technology. Implementing sublime interfaces between technology and people will break down ambiguous black-box models. Besides transparency, other ethical issues will arise and one should discover and take this into account as early as possible in the design process.
- Dickson, B. (May 12, 2017), What is Narrow, General and Super Artificial Intelligence, TechTalks
- Handley, L. (November 19, 2018), The 'world's first' A.I. news anchor has gone live in China, CNBC
- de Almeida, L. (2018), The Difficult Task of Orchestrating AI-powered Services, Touchpoint, vol 10, no 2: (26-30)
- De Ketelaere M. (29 November, 2018), Gent, Trefdag Digitaal Vlaanderen, Van AI tot IA
- Borgefalk G. (2018), Designing Relationships with Technologies which pass as humans, Touchpoint, vol 10, no 2: (36-39)
- Kozubaev S. (2018), The future of service design in a post-human world, Touchpoint, vol 10, no 2: (44-47)
- VanAntwerp S. & Mhanna S. (2018), Human Machine collaboration: Designing for a new kind of relationship, Touchpoint, vol 10, no 2: (14-17)
- Ethics of Artificial Intelligence, Wikipedia, retrieved January 10, 2018.
- Three laws of Robotics, Wikipedia, retrieved January 10, 2018.