Research on Intelligent Design Tools to Stimulate Creative Thinking

1.1 The Popularity of Online Collaboration with Non-designers’ Teams

Nowadays, collaboration among multiple people, teams, and even companies has been a common situation. Daily meetings, regular project discussions, progress reports, etc. are common in everyday collaborative activities, and with the development of network technology, online collaboration has become a new form of teamwork, but it is also difficult to make sure that everyone is online in real time every day. The significance of collaboration tools is actually here, helping everyone to communicate seamlessly whenever they have requests, sharing information, and updating progress.

Although different teams have natural differences in terms of number, location, resources, goals, processes, etc., there are similarities in their division of design tasks and daily needs.

1.2 The Design Process is not yet Universal

A correct and reasonable design process can guide the team to design efficiently, but the design process hasn’t been widely used at present.

The design process is a set of problem-solving methods formed by the induction and modeling of the designer’s thinking. It usually includes a series of explicit steps to provide participants with a thinking frame, which plays an important leading role. Under the guidance of the design process, even non-designers can quickly grasp the design method, so that all staff can participate in the design work in an orderly manner. However, at present, the design process has not been widely used in design work, and non-designers’ teams need the guidance of correct and reasonable design process.

1.3 The Trend of Using Artificial Intelligence in Daily Work

We are already in the digital world. Computers are increasingly entering people’s work and life. Driven by efficiency, value creation and informatization, artificial intelligence has become a hot topic in technological development. Big data and deep learning technologies also provide important technical support for artificial intelligence. The way of value creation in the future will be efficient information processing, and artificial intelligence will be a powerful force for the development of technology and society. Its historical significance is no less than that of the industrial revolution [1].

2.1 Design Thinking and Design Process

Design thinking has attracted widespread attention since it was first cited by Peter Rowe, Dean of the Harvard School of Design in 1987 [2]. The use of design thinking can encourage teams to come up with ideas democratically, thereby improving the efficiency of communication between participants, proposing and selecting solutions that are suitable for each participant. The design thinking has also been verified by practice since its inception, which has promoted design positively. The trump majors of top universities in the world such as Harvard University, Stanford University and MIT, the project of Fujitsu Japan called Design Thinking for Future Schools, the “Design Thinking Toolkit” jointly launched by the well-known design company IDEO and Riverdale Country School which is in New York, are all typical examples of the use of design thinking [3].

During the development of design thinking, many models have been proposed [4]. The earliest process expression of design thinking was almost a repetition of the traditional design process. Later, deeper empathy and more specific forms of multidisciplinary collaboration were added. The design process proposed in Herbert Simon’s 1969 book The Sciences of the Artificial [5] is definition, research, creativity, prototype, selection, implementation and learning, which has become the cornerstone of the design process. Professor Liu Guanzhong of Tsinghua University divided the process of design into six steps, which includes observation, analysis, induction, association, creation and evaluation, progressively building up the model of problem solving [6].

The book A Designing for Growth: A Design Thinking Tool Kit for Managers [7] written by Jeanne Liedtka, professor of business administration at the University of Virginia’s Darden College, and Tim Ogilvie, founder of innovation consulting firm Peer Insight, redefines design process to 4 W, that is, design can be seen as the answer to four questions, which are What is? What if? What wows? What works? (Fig. 1).

The process of 4W

Full size image

The British Design Association’s double diamond model uses 4D to define the process of design thinking, including Discover, Define, Develop and Deliver. The double diamond chart divides thinking into two phases, namely the divergent phase and the contractive phase. The core is to find the right problems and finding the right solutions. This is a structured design method that is loved and used by many designers (Fig. 2).

Double diamond model

Full size image

The process used by IDEO includes three stages: inspiration, conception, implementation. It also released a set of design method cards [8], covering learning, observation, inquiry and so on. IDEO also developed an environment-based toolkit [9] and the HCD process [10] (human-centered design) was reinterpreted as an acronym for Hear, Create and Deliver.

D.school [11] defines design thinking as a five-stage process, which includes empathize, define, ideate, prototype and test. Each stage contains different goals, implementation principles and specific method tools. It shows that design thinking is an iterative process rather than a linear process (Fig. 3).

The design thinking process of d.school.

Full size image

Design Sprint is a rapid design method proposed by Google. It is mainly applicable to team collaboration tasks that need to come up with design plans in a short time. It includes 6 stages: understanding, definition, divergence, decision, prototype, and verification. This approach allows non-technical designers to understand the methods of designing quickly and promote people with different backgrounds to participate in collaboration.

In a word, although there are many types of models about design and the theory of design thinking is diverse and complex, when peeling off the surface to explore the essence, we can find that its core connotations are basically the same. Based on the model discussed above, we take the double diamond model as the basis and incorporate the concept of design thinking into further exploration of the design system.

2.2 Design Tools

In order to explore the role that AI can play in design systems, we conducted research on traditional design tools. For example, the analysis method called SWOT [12] is used to formulate corporate strategies and analyze competitors in the early strategic planning of product. It integrates and summarizes internal and external conditions from various aspects and analyzes the advantages and disadvantages, opportunities and threats, positioning the product and analyzing the core competitiveness of it among similar products. For another example, Japanese professor Noriaki Kano invented the Kano model, which is a useful tool for classifying and prioritizing user needs, analyzing the impact of user needs on user satisfaction and reflecting the non-linear relationship between product performance and user satisfaction. In the kano model, the quality characteristics of products and services are divided into five types: Basic Quality, Performance Quality, Attractive Quality, Indifferent Quality, Reverse Quality. It classifies user needs and prioritizes multiple function points and can be used during the convergence phase of the design process. Design tools can be divided into figurative and abstract tools, traditional tools and new tools [13]. Material tools are those such as pens, paper, whiteboards, pictures, which are also traditional tools. Abstract tools are those such as words, conceptual framework, theory, brainstorming. AI assistive tools are also new forms of design tools.

2.3 Design Systems and Collaboration Platforms

There exist some systems of design and collaboration. For example, Teambition [14] is a collaborative creation tool for teams. InVision [15] is a prototype design and team collaboration tool. UXPin [16] is a collaboration tool, its team version can add three different member roles into a project, including administrator, creator, and collaborator. Figma [17] is real-time collaborative UI design tool based on browser and cloud, allowing multiple designers to edit a project online at the same time. Zeplin [18] is a standardized design style tool for designers and developers to work together.

However, there are some shortcomings. According to the inspection of these systems, we understand the current situation of this type of systems: firstly, These systems are not good at guiding the design process and the function of collaboration is greater than the function of design assistance. Secondly, AI assistance has not been widely used. Thirdly, these systems are highly specialized and not suitable for non-design professionals.

2.4 AI Technology

AI assistance is one of the focus of our research due to the broad application prospects of AI in recent years. Artificial intelligence technology was first formed in the middle of the last century. In recent years, with the development of cloud computing, intelligent algorithms and other technologies, huge improvements have been made in computing capabilities, data analysis and practical applications [19].

The theory of design process and design methods have already been mature, but the design system only stays as a tool and cannot guide the design well. At the same time, the AI technology which is getting more and more mature has not participated in the design process. We think this is a great opportunity.

3.1 Concept

We hope to construct a design system with the participation of AI, so We need more detailed research on the role of AI in designing systems. In this article, we preliminarily propose a design tool that can be combined with the theory of design process and based on the concept of AI.

We gave it four characteristics:

1. 1.

   Guidance: Design thinking and design process are integrated into the design system, and users are guided by the system to do design tasks based on scientific thinking and process. A scientific and reasonable design process can effectively lead the divergence and convergence of design thinking. We hope that new design tools can use the design process to play the role of guidance for participants.

2. 2.

   Intelligence: We hope we can combine this system with artificial intelligence and big data. The new design tools we propose will take AI as the core which will intelligently give tips and help in the design process according to the pace of the designers’ discussions.

3. 3.

   Non-interference: Our design tools will not directly interfere with the design process, but will give tips and help during the design process. In this way, it will not unduly interfere with the original design process and the behavior of participants can be observed and the system can be improved more effectively.

4. 4.

   Modularity: The design process is complicated, but if it is decomposed, the design work can be simplified. Therefore, we need to decompose the design system into modules one by one from top to bottom, and each module implements a subfunction. Modules can be combined or decomposed, and they are independent relatively as well as interacting with each other at the same time. By completing the tasks of each module, the entire design is completed step by step, making the design tasks easy to control and manage.

3.2 Design

The target users of our design are college students who don’t major in design. The common characteristics of these students are that they are less familiar with the design process and are often not good at expanding ideas and participating in practice compared with students who have learnt about design.

The final product is a design thinking extension assistant. When students discuss, it will provide students with information that can help them expand their thinking according to the pace of their discussions, eliminating the limit of students’ thinking, intelligently controlling the rhythm to better guide the design process. Since the use of images and texts can help people process complex information, our mind extension assistants will provide information in the form of images combined with words [20].

In order to construct this tool, we need to explore the relationship between tool forms and design process. Tools can help students expand their thinking by giving materials, which can be images or texts. The characteristic of images is that it is able to elicit richer and deeper associations, while the texts are characterized by easier understanding and higher efficiency. In the design process, design thinking can be divided into two parts: divergence and convergence. We use ​​the double diamond model [21] to conduct experiments to explore the applicability of different forms of design tool in different phases of the design process.

3.3 Prototype

We made simple prototypes with paper and do experiments. We have two goals in the experiments. First, during the phase of thinking convergence and divergence, which method performs better, and which is more suitable for the entire design process. The second is to discuss the symbiotic relationship [22] between the new design tools with AI’s intervention and users, and the relationship between the AI and users. In our prototype, we developed conservative and passive strategies for AI to test the relation.

Our experiments need to collect a lot of pictures and texts. In order to simulate AI’s decisions, we collected it manually. At the same time, in order to avoid the interference of subjective factors, the task of collecting materials was distributed to different types of people. Each person collected 10 words/phrases and 10 pictures for the specified requirements formulated for the experiment. We collected a total of 200 words/phrases and 200 pictures (Fig. 4 and Fig. 5).

Partial pictures

Full size image

Partial words/phrases

Full size image

4.1 Test

Experimental method: Wizard of OZ. Wizard of Oz (WoZ) is a technique for prototyping and experimenting dynamically with a system’s performance that uses a human in the design loop. It was originally developed by HCI researchers in the area of speech and natural language interfaces as a means to understand how to design systems before the underlying speech recognition or response generation systems were mature [23]. We use the WoZ method to simulate the role that AI can play in the design process. We conduct experiments by simulating actual discussion scenarios to collect various feedback during the design process.

Staff assignments: We did 6 groups of experiments, each group had 4 students, and one of the students played the role of AI to participate in the discussion.

Experimental process: The discussion lasted for 1.5 h each time. In the discussion, we introduced the concept of the double diamond model [21]. There were 2 periods of divergence and 2 periods of convergence. The processes include demand exploration, demand definition, brainstorming, and solution derivation (Fig. 6 and Fig. 7).

Experimental process

Full size image

The phases of the experiment

Full size image

Experimental design: In the experiment, we gave 3 groups of students picture materials and 3 groups of students text materials. “AI” provided participants with information in two forms: passive and active. Passive means that “AI” will give students relevant materials for discussion every 30 s. Active refers to the fact that during the discussion, “AI” will determine whether the frequency of giving information should be increased according to the trend of the discussion.

4.2 Result

Experiments showed that AI’ intervention in the design process was helpful. We observed 6 groups of discussions. In the 1.5-h discussion, the stage of thinking divergence took 65 min and the stage of thinking convergence took 25 min.

In general, students were more dependent on the “AI” assistant when their mind diverged and were less dependent on the “AI” assistant when the mind was shrinking. In the discussions of 6 groups, “AI” gave information for a total of 1,332 (180 * 6 + 252) times, of which 960 (127 * 6 + 198) times were given when the mind was diverged, and 372 (53 * 6 + 54) times were given when the mind shrank. When the mind diverged, 28.9% of the materials proposed were discussed, while only 9.7% of the materials were discussed when the mind was contracted. Therefore, we think that this “AI” assistant is more suitable for the divergent thinking stage in the design process (Fig. 8).

Divergent thinking and convergent thinking

Full size image

We use the 45-min divergence of thinking in the experiment as an entry point to explore the performance of texts and pictures in the discussion. On the whole, the performance of the pictures was better. In the three discussion groups that provided texts, “AI” gave materials for a total of 504 times, of which only 24% (121) were used in the discussion, and only 30% of the texts used were converted into specific ideas (11). In the three discussion groups that provided pictures, AI gave materials for a total of 456 times, of which 34.4% (157) were included in the discussion, and 14% (64) of the pictures used were converted into specific ideas (Fig. 9 and Fig. 10).

Text date in divergent thinking

Full size image

Picture data in divergent thinking

Full size image

Let’s take one of the groups given texts and one of the groups given pictures as an example to explore the difference between when “AI” is in active and in passive states. When “AI” was passive, pictures were more useful than texts. Both groups received materials for 127 times, of which 44 were pictures and 24 were texts. Through observation, we can easily judge that the picture materials are more likely to cause students’ association, and the content of discussion will be deeper and more specific. However, when “AI” was active, the texts’ guidance was stronger. “AI” provided 39 words/phrases to one group, of which 18 were used in the discussion, and 25 pictures were provided to the other, of which 10 were used in the discussion (Fig. 11).

Active vs passive

Full size image

Through data analysis and simple interviews with some students after the experiments, we can draw the following conclusions about this study:

1. 1.

   At the stage of divergent thinking, students rely more on AI thinking assistants. At this stage, it is difficult for students who have not studied design methods to start discussions, and the tips given by the AI ​​assistant can help them start the topic. However, at the stage of thinking convergence, students basically draw conclusions from previously divergent ideas without much help from thinking assistants.

2. 2.

   Pictures are more helpful for divergent thinking. The texts are too abstract. Although it is easy to understand, students generally think that the texts cannot trigger more associations. Although the picture materials need to be interpreted, they can help students trigger a series of chained associations, which is more helpful for divergent discussions.

3. 3.

   If the relationship between the AI and the user is different, the behavior of the AI will also be different. When AI is passive, the pictures are more effective and when AI is active, the texts are more effective. Through interviews with several students, we found that this is because the pictures take a certain amount of time to interpret, so if AI is active, it occupies the main role and will interrupt the students’ thinking. At this time, the characteristics of texts are more efficient, the continuous output of which can help students compose scenes and enrich the imagination. when the AI is in a passive state, the role of AI is weakened, and students’ discussions are dominant, and the pictures can stimulate deeper discussions.

4. 4.

   AI should not overly interfere with the design process, but it is more suitable for AI to play a role of supporting design. In the experiments about the design process involving AI in this article, we just provided simple materials without other intervention, which has already affected the discussion process of students to some extent. Therefore, we think that when AI participates in the design process, it should not interfere too much with the students ‘ discussions, so as not to affect the students’ design thinking too much.

In general, the combination of AI and design processes will undoubtedly become an important trend in the future. In order to meet this trend, we will continue to explore the possibilities of future design tools. In the future, we need to conduct advance research on the form and intensity of AI’s intervention. We will explore how AI is involved in other design processes in more ways in the design process. Since this article mainly discusses the intervention of AI in design thinking, in the future we will study how AI can better participate in the process of thinking convergence. And we will gradually form a more complete AI design system concept through several experiments.