The underlying and initial question to this thesis is: “How might we design interactions for New and No User Interfaces?”. To answer it, this project has covered three main areas:
The age of conversational UIs in relation to the current state of the industry
Principles and guidelines for designing interactions and experiences for conversational interfaces
A real world case study as an exemplary exploration of the process of designing a conversational interface
This final chapter reviews what has been covered throughout this study, concludes, reflects and looks ahead trying to find implications this work could have or potential fields where further research could be carried out.
8.1. The state of the industry and the age of conversational UIs
“How might we design interactions for New and No User Interfaces?” serves as a point of departure for this study. The question closely relates to conversational UIs as a new form of visual or nonvisual interfaces. The advantage of conversational interfaces is a smaller learning curve, as the most human form of interaction is written or spoken language. The rise and popularity of messaging services and smart assistances comes across as a stop-over in the digital evolution. “The rise in popularity of these apps recently brought me to a startling observation: Advances in technology, especially in AI, are increasingly making traditional UI irrelevant (Aube and TechCrunch, 2015)”. Services such as the Google Assistant or Apple Siri move from visual to audio interfaces. Additionally, concepts such as “Anticipatory Design” have gained momentum. “The Internet has given us an abundance of choice and an abundance of information to inform those choices. The result is that our lives are burdened with approximately 35,000 decisions a day”, Shapiro (2015). Krishna (2015) adds the more choices a user has, the harder it is to make one.
Rather than actively interacting with devices and interfaces, Shapiro (2015) believes the next big breakthrough in design and technology will result in products, services, and experiences that eliminate needless choices from our lives. Instead, they will make decisions on our behalf, freeing us up for decisions we care about (Shapiro, 2015). Soegaard, M. and Interaction Design Foundation (2016) state that a core goal of UX is to make choices and the life of users as easy and painless as possible. This aligns with Case (2015) who suggests it is not the goal to help the user make a decision but to create an ecosystem where decisions happen automatically and without user input (Shapiro, 2015).
Such systems could eliminate the need for active user interaction and make screen interface redundant. Krishna (2015) believe the best result for any technology is to solve important problems in impactful ways. This leads to his four principles:
The best design reduces work.
The best computer is unseen.
The best interaction is natural.
The best interface is no interface.
Current conversational applications and services, especially voice based, seem to aim in this direction. Nevertheless, and how this research fails to consider is the circumstance, that conversational UIs are only one form of a New or No User Interface. Relating back to the original research question, New and No User Interfaces might occur in forms we are not able to think of yet.
Others forms are already here but have not been mentioned in the scope of this study. There are trends and new technologies on the verge of breaking through to mainstream audiences. Augmented and virtual realities offer new interfaces or extend existing ones. Other applications are smart IoT-devices. Any device might become smart or anticipatory and offer a unique interface humans interact with, e.g. through physical gestures or movements or even pure imagination.
However, visual interfaces are far from redundancy. In spite the fact there is a movement towards screenless interfaces, screens still serve many purposes. We are visual beings. Apart from the move towards enhanced visual experiences via AR or VR we rely on visual stimuli in various aspects of our life. Videos and photos are intensively used types of media and will not go away anytime soon. The choice of an interface closely relates to the goal or purpose of a product and its user.
Other aspects that have not been covered and leave room for further exploration when it comes to intelligent services are of a moral nature. Smart systems rely on data. They need to know their users, their traits, habits, preferences and behaviour to provide value. This comes at the cost.
Users need to provide their data to the provider of a service. This raises questions about data privacy, intimacy, data protection or security. Who can access data, what is it used for and what happens if data gets lost, stolen, corrupted or abused? This study does not cover these aspects. But their handling when it comes to providing use and benefits to humanity, in the long run, are crucial. Further research could be carried out in these fields in order to answer questions such as “How can personalised and anticipatory services and products maximise benefits for users without depriving them of their personality?”
8.2. Principles and guidelines – Designing with the human in mind
Part of the initial question is “how to design” for new and no user interface. As stated earlier, the term “New and No User Interfaces” is broad and it is impossible to cover every possible form of interface and interaction that is currently available or that might be available in the future. “No matter the client, the industry, or the problem we are trying to solve, the creative process for developing technological solutions repeats the lazy act of drawing a screen (Krishna, 2015)”. To avoid this trap, the focus has been put on conversational UIs that may be visual or nonvisual.
To acknowledge an ever-changing technological landscape, new devices and new interfaces, a variety of principles and guidelines have been examined. From established and general design and UX guidelines to newer and more specific guidelines for conversational UIs. Especially UX guidelines focus on making products usable, reducing work for the user and getting jobs done. They aim at improving usability. Nevertheless, design is not all about usability. Norman (2015) adds visceral and reflective layers. I think this makes sense for various reasons.
First of all, products and services do not exclusively provide use. A core principle when designing is not only knowing the user but knowing what a product’s or service’s goal is. Furthermore, and as Rams’ (no date) states “good design is long-lasting”. Although his approach is based on his personal opinion and argumentation mainly lies with the visual appeal that should stand the test of time, this aspect also bears similarities when it comes to the impact a product has. It relates to Norman’s (2015) reflective level of design and acknowledges that good design is more than just looks and usability. Good design creates value and pleases beyond its visual appeal and ease of use. Good design builds a relation with its user. Good design provokes a long-lasting emotional reaction and triggers a reflection that can be turned into stories to tell and share and that can even be debated upon.
The design principles developed in the scope of this study attempt to account for this. They should provide scaffolds upon which, New and No User Interfaces may be designed. Despite the fact that they provide a foundation for designers, they fail at providing concrete and hands-on advice and tools a designer could apply right away. To do so further exploration and research would have to go into specific products, services, scenarios, use cases and interfaces.
Related to this topic a discussion arose around the term “experiences” when discussing the outcome of a user testing and applying design guidelines. Based on the testings results I argued “hard facts” such as contact details would have to be visible and available at first sight. Hinderling (2016) opposed to this by advocating a more minimalistic and clutter free layout through “progressive disclosure”. Although agreeing with this approach, I pledged for the “user experience”. An argument used by Hinderling (2016), too. As it turned out during our debate, Hinderling (2016) referred to the “visual” aspect and appeal and thus the “visual experience” of the product whereas I talked about the “behavioural layer” relating to usability based on Norman (2015). This was a valuable example of how similar terminologies may be interpreted differently. Openly debating helped to shed light on the different positions. In the end, a visually appealing solution that also catered for basic user needs was found.
When designing for complex systems we are faced with today, knowledge and skills in various disciplines are required. When creating a conversational UI, established UX, web- and interaction design principles and techniques should not be dismissed as this study has shown. Within this study, findings from secondary research have been applied to industry collaborations. At the same time, these collaborations provided validation and new insights for setting up the design guidelines presented. User testing revealed how applying these principles positively resonated with the users. More extensive testings and different cases studies and applications would be necessary to validate further, and questions remain?
How might we differentiate between different terminologies such as guidelines, principles, recommendations or rules?
How might we constantly assess and validate appropriate guidelines for specific scenarios?
How might we make best use of existing guidelines?
How might we appropriately adapt and apply new guidelines for new challenges?
8.3. The process of designing a conversational interface
In the scope of this study, a user-centred approach was applied to a real world project, in which the goal was to build a conversational UI in the form of chatbots for an agency website. Several challenges occurred during this process and can be grouped into two categories: 1. Project management. 2. Tools and artifacts.
8.3.1. Project management
Applying an agile sprint approach adopted from Jongerius and Berghuis (2013) in conjunction with the Revamped Double Diamond (Nessler, 2016) worked but revealed some drawbacks. When working on a real-world project, dependencies come into play. Human resources assigned to the project were sometimes limited, pulled off or not available. This resulted in delays and vacuums in which, the planned tasks could not be carried out and thus had to be postponed, rescheduled or abandoned. This was not always ideal, but it is a reality when dealing with such a setup. One way to deal with it was to insist on regular meet-ups and openly discuss the project status, challenges and thus manage expectations. This sometimes resulted in redefining the scope of a prototype, other deliverables and deadlines. The launch of the MVP was planned but had been pushed back.
Applying a user-centred approach can be challenging when the environment and processes are not necessarily set up for it. To my advantage, I dealt with an internal project, and the management was open to suggestions and when it came to applying them.
I would consider applying an identical process in other projects again. To apply the learnings, though, holding a position with corresponding authority over resources or ownership in a project, team or company, and an open-minded work-environment seem essential for success.
How might we best incorporate a HCD process in day-to-day business?
How might we assess the best project management approaches for specific projects?
How might we deal with changing circumstances that undermine a specific process?
8.3.2. Tools and artifacts
Relying on established UX tools and frameworks worked. In some cases, existing tools had to be adopted for their particular use as shown with the bot persona example.
It needs to be pointed out that there have been various challenges and a certain amount of complexity when it comes to logic, databases, dialogues and the visual execution. Compared to large-scale intelligent systems relying on big data, the complexity of our project was relatively low. More complex projects would require a larger infrastructure, more resources and more diverse knowledge within the team. Therefore, this case study only provides a glimpse of what a corresponding process of developing a conversational UI on a larger and more complex scale might look like, but it fails to verify whether the tools and deliverables could be scaled up.
This also relates to a question that has not been entirely answered. How can we collaboratively prototype and visualise conversations to make them comprehensible and testable? Different attempts at using flow-charts, mind-maps, spreadsheets and conversational prototypes had been made. An entirely satisfying solution could not be found, though.
There was a debate around low vs. high-fidelity prototypes. I advocated for low-fidelity prototypes and quick user testings. Hinderling (2016) argued for high-fidelity prototype testings, especially when testing with external stakeholders. Hinderling Volkart is a design savvy company that is renowned for its visual output, and they value it (Hinderling, 2016). Relying on higher fidelity prototypes generated dependencies on the code and design side that I could not entirely compensate for. In retrospect, I could have evaluated other ways of testing conversations, which I failed to so due to setting other priorities.
Overall, tools such as scenarios, personas and visualisations were beneficial during this process. At the same time, the project would have gone forward without applying these artifacts I suppose. Whether the process would have been different may only be assumed. At this point some questions and potential topics for further research remain:
How might we evaluate and apply appropriate tools to visualise and prototype conversations collaboratively and test them?
How might we define and apply a human centred and UX toolkit and terminologies in a constantly evolving technological landscape?
How might we justify the use of such tools in the industry?
How might we create a shared understanding when it comes to design terminologies?
8.4. Final thoughts
Overall, applying the Revamped Double Diamond approach to solving the initial research question helped the process, provided structure and guidance. Nevertheless, various obstacles occurred. First of all, the initial question was extremely broad. This resulted in challenges, especially during the first research phase. Looking into one particular topic caused other sub-topics or related topics to pop up. And even though research is a divergent phase during a project, the aspect of getting lost in specific details was constantly present.
Entering a collaboration with a clear goal helped to synthesise and frame the angle of the project. Nonetheless, new challenges occurred in this context. Collaborating with a partner also meant complying with their agenda, resulting in dependencies and less flexibility. Being confined in this also meant personal goals regarding progress and process could not always be met. As initially planned an MVP should have gone live before the deadline of this research project. At the point of this writing, the MVP is not yet ready for release. It would have been valuable to be able to test the final MVP, evaluate and analyse the findings and verify the design principles set up in this project.
A main challenge within the overall research project was the volatile nature of this topic, the lack of academic literature and the rapid progress in the field. Throughout the entire process and up to this day, new articles, guidelines, principles, products and services have been introduced. Thus, new aspects opening new perspectives have constantly appeared but had to be left out due to time constraints.
On the design side, I have benefited from design theory and existing research in the field. When debating aspects of design, it seems valuable to be aware of, and familiar with this theoretical background. The three layers of design advocated by Norman (2015) provide helpful guidance when applying, assessing or debating design. At the same time, I need to question the guidelines I have set up within this study by questioning the overall meaning of “design”. What is design and how does design come into play when “designing” interactions and experiences for New and No User Interfaces? Huge Inc. and van Hoof (2016) proclaim designers need to have skills and knowledge in UX, emotional awareness, statistics, psychology, linguistics and other fields once AI comes into play.
I would personally love to have knowledge and expertise in all of these fields, but is this even possible for one human being? I highly doubt it unless we create the “uber-human” that would have to rely on some intelligence enhancement system itself. In a world of ever-growing complexity, a broad knowledge is gold, assuring that one may comprehend challenges, relations and interconnections between different fields. Expertise is silver and valuable in at least one or even better two or more fields. In the end, it comes down to the ability to collaborate with others and uniting the right people for the right job.
This said the overall topic leaves room for more exploration and research in various fields. Some explanatory and more holistic questions for future research may be the following:
How might we define the role of a designer or a UX designer today and in the future?
How might we define the role of design in general and when can we label something “design”?
- How might we collaboratively work together and apply individual skills when needed?