As designers, we worked hard to apply the best designs principles and justify how design decisions are made through phases, methods, or tools.

So, where do we stand as designers when artificial intelligence (AI) takes over the design process and performs better than us when it comes to delivering the ultimate user experience?.

Our limitations as designers.

Robeto Verganti, Luca Vendraminell, and Marco Iansiti highlighted that there are three limitations that designers face in the age of AI:

1. Scale and people centeredness
2. Scope
3. Learning and iteration

Each of these limitations are detailed below.

1. Scale and people centeredness

Traditional design approach

We always have a specific target of users to design for and manufacture at scale for use once a solution is tested successfully.

The reason why we follow such a cycle is that it is not economically feasible, especially in time and resources, to design a different solution for every user.

A product will only be redesigned when there are new insights obtained from users and when the marginal value of the product supersedes the cost of design. The cycle kick-starts again when a significant effort and investment are made by the organization. There is no doubt the design process is labor-intensive.

The problem poses in this cycle as shown in the diagram below is that innovation happens episodically in lumps. By the time, designers are prototyping a solution to feed into development of the future version of the product, the solution becomes rapidly old because of the gradual change in the context of use by users.

AI factories approach

AI removes the limitations of scale and people-centeredness by leveraging real-time data on each individual user interaction and develops a specific solution immediately for the users. This is doneby performing deep learning on algorithms and data sets. In the design context of AI factories as shown in the diagram below, the design, delivery, and use phase happen simultaneously with zero time and cost.

2. Scope

Traditional design approach

Products are usually designed for a specific industry with a target user. Once it is released, it is unlikely that it applies in a different context. Hence, it is only confined to a specific project.

For example, a car is a mode of transport. Moving from there to entertainment services is unlikely.

Once a design brief is defined, the creativity happens only within the space of the brief.

AI factories approach

The design brief is fluid and reframed even when a product is released in AI factories. Making it easier to imagine radically new services.

3. Learning and iterations

Traditional design approach

Based on the traditional design context, it takes time to move from one phase to another where just as designers are making sense from the observations of the real use of the product, users’ contexts gradually evolve and solutions coming out from designers are rapidly becoming old.

AI factories approach

Referring to the diagram above on design context of AI factories explains that each time a user use, it activates the problem-solving loop (PSL).

PSL leverages data obtained from users and directs algorithms to directly learn towards improvements.

The learnings conducted by machines are based on real use rather than testing prototypes in a simplified context.

Learnings have become more people-centered in this case as data comes from the early use by the same person instead of leveraging insights from other people using previous generation products.

Every user interaction now offers an opportunity to the machines to conduct new experiments with different logic.

Therefore, many development decisions are now made through PSL because it is :

• autonomous
• human capital free
• easy to scale without redesign
• provides a variety of data without a large investment in research and development
• provides solution targeted to individual person

Netflix currently also makes decisions about which movie to show, how to display them, which picture to represent them with and other design decisions by using algorithms under PSL.

With AI, the specific solution experienced by an individual user (what she sees on screens) is not only delivered but also designed by PSL.

How disruptive is this change for designers?

The design process

AI algorithms do not reason like humans, where they do not replicate the thinking of an engineer or a designer. The algorithms work differently entirely where we cannot fully understand how it connects pieces of information, what it emphasizes or what it omits.

As the objects of the design changes from designing solutions to PSL, there is no doubt that the process of the design (the “how” of design changes as well).

How does one apply concepts such as incremental or radical innovation in a context in which solutions keep evolving?

Based on Niya Stoimenova and Rebecca Price findings in exploring the Nuances of Designing (with/for) Artificial Intelligence, AI is becoming less like a tool and more of an extended mental apparatus of a person which plays profound ethical implications in our society.

Design is now transitioning into a more critical role of developing an ethical framework based on requirements for all potential users, not just a lead user or set of users.

Accountability on decision-making

Increasing adoption of PSL in design means we are also increasingly abdicating our power to make decisions based on our judgments, which includes moral convictions.

Given the speed and precision of problems solved by algorithms, we tend to accept these answers easily without knowing that these machines can produce irreversible situations.

In Christine Moser’s article on What Humans Lose When We Let AI Decide discusses the importance of employing judgment that encompasses an intrinsic moral dimension which relies on:

• Reasoning
• Imagination
• Reflection
• Examination
• Valuation
• Empathy

in order to form our decision that accounts social, historical context, and different possible outcomes when we get AI to work for us.

How to move forward? Choosing between the inaction or action

Inaction

If AI poses such a threat to control human freedom by disregarding human innovation, ambiguity, and imperfection, maybe the way to move forward is to stop trying to automate everything that can be automated just because it is technically feasible.

Evegeny Morozov describes in his book To Save Everything, Click Here, that humans have lost basic capacity of moral reasoning with the lackluster of institutions that don’t take risks and only care about the financial bottom line and perfectly control social environment.

If the idea of perfect is what we want that can be met by AI, why would anyone bother to innovate?. Should we just stop automating ?.

Action

We need to learn how to work along with AI by understanding the use of technical images or data that is feeding into the algorithms, such as how to design for PSL.

In practice, we need to judge case by case basis whether, how, and why to use AI.

If AI is adopted, we need to remain vigilant and doubt the outcomes as our inflated expectations of what algorithms can do, and their actual capabilities can potentially bring harmful consequences on society that violates moral justice and unfairness.

Questions yet to be answered

There are still so many questions yet to be answered on whether AI should be adopted, such as :

• Is it appropriate in any context? Does it depend on industry or company-specific factors, including strategy or culture?
• How might these products, services, and systems grow with users in surprising and delightful ways over time without inflicting harm when AI is implemented?
• How does AI induce practice challenge design fundamentals? Is it still going to be user-centered ?.

References:

• Stoimenova, Niya & Price, Rebecca. (2020). Exploring the Nuances of Designing (with/for) Artificial Intelligence. Design Issues. 36. 45–55. 10.1162/desi_a_00613.
• Verganti, Roberto & Vendraminelli, Luca & Iansiti, Marco. (2020). Innovation and Design in the Age of Artificial Intelligence. Journal of Product Innovation Management. 37. 10.1111/jpim.12523.
• https://sloanreview.mit.edu/article/what-humans-lose-when-we-let-ai-decide/
• Morozov, Evgeny. 2013. To save everything, click here: technology, solutionism, and the urge to fix problems that don’t exist.