Keywords

1 Introduction

Human-Computer Interaction (HCI) is a field that studies the design and use of computer technology, especially the interaction between computers and people. Recently, several segments of the Brazilian industry have invested in HCI activities by hiring in-house experts and adjusting their business processes [1]. However, the low maturity in HCI, organizational culture, and the growing popularity of Agile Software Development (ASD) have reduced design time and decreased the odds of conducting research with end users [2].

To deal with these obstacles, many design practitioners have adopted solution-focused approaches aimed at synthesis. These approaches promote the definition and construction of solutions in short design cycles, increasing the understanding of the problem through the evaluation of candidate solutions. They are mostly adaptations to design approaches that leverage synthesis, analysis and evaluation [3, 4].

In 2007, Aureliano proposed eXCeeD – an adaptation of a Communication-centered Design approach that combined eXtreme Programming (XP) values with a streamlined HCI design process grounded in Semiotic Engineering (SemEng).

Communication-centered design approaches emerged as “an attempt to ensure that the domain concepts to be communicated to users are well represented and understood by every team member before proceeding to later design stages” [5].

In this approach, the search for understanding is supported by expressions used by the users when they interact with help systems (e.g.: “How do I do this?”) [5]. Knowledge representation happens by building models that represent the interaction as a conversation between users and the system [6].

This study aimed to adapt eXCeeD, following its proponents’ recommendations, and to contribute and increase the understanding of the communication-centered process. In order to achieve this goal, we observed the process adoption on a project in a corporate software development environment.

Among the suggested changes to eXCeeD, we can list: properly enabling the participants to carry out the activities, including an initial stage dedicated to user research and task analysis, besides investigating the explicit use of eXCeeD’s questions to support semi-structured interviews with users and the study of the adoption of interaction models and regular notes.

Among many challenges of a study like this, we focus our discussion on the differences between a process like LeanCCD as proposed and how it is followed in practice. This discussion motivates us to reflect on how methodical (or not) such a process should be [7]. It also invites us to reflect on the designers’ difficulties to follow prescribed procedures and to adopt specific techniques, particularly when they are not related to their previous experiences [8].

In the next section, we briefly present and compare the communication-centered approaches to interactive software design. In the third section, we present LeanCCD by detailing the adjustments made to eXCeeD. Next, we describe the case study that applied LeanCCD and a summary of the findings. Finally, we discuss some of these findings, concluding with some remarks and directions for future work.

2 Communication-Centered Design Processes

The CCD authors believe that in order to design the human-computer interaction, it is necessary to facilitate communication among designers to create a shared understanding of the domain and how the application should support users in that domain [9]. Communication-centered processes make use of questions used to build help systems [10] in order to capture this knowledge. On the other hand, design is achieved using MoLIC interaction models, among other representations.

A major contribution of SemEng to the HCI field was the proposal of a communicability evaluation method, representing breakdowns in the user-system communication (for instance, “I can do otherwise.”). Silveira [10] used them to define expressions to access help systems (for instance, “Is there another way to do this?”). These expressions evolved in eXCeeD for questions that assist semi-structured interviews with users (for instance, “In what other ways would you {need/like} to do this?”).

With regard to the models, MoLIC has been used as a shared representation with two distinct goals: to represent the communication between the users and the designer (represented by the user interface, at interaction time) and to promote communication among members of the design team (at design time). The language proposes the creation of three representations: a model of user goals, an interaction model, and a conceptual sign schema.

Studies with MoLIC form the basis for the knowledge related to the adoption of CCD. eXCeeD, assuming that such an approach would generate extensive documentation and in order to adapt it to the development using eXtreme Programming [11], distributed activities, initially arranged in a waterfall model in the CCD, into design and development cycles.

Each cycle starts with prioritizing and detailing a user goal or a set of user’s goals. Then, the designers define the user interface and the system interaction through a combination of interaction models and user interface paper sketches. Finally, they evaluate the candidate solution with end users, using a technique called paper prototyping [12]. Figure 1 shows a summary of eXCeeD.

Fig. 1.
figure 1

Graphic summary from LeanCCD’s steps and activities

Full size image

A case study evaluated the adoption of eXCeeD on a small project in an academic environment. The results pointed out a lack of understanding of the users’ needs during the description of the user goals, and the incorrect application of paper prototyping as defined by Snyder [12]. During interaction and interface modeling, disagreement among designers have resulted in the abandonment of both MoLIC diagrams and regular notes, weakening the conclusions regarding to the light documentation maintenance proposed by the process. Throughout the study, the designers did not use the questions explicitly, although subsequent tests have shown that designers addressed many of them intuitively.

3 Lean Communication-Centered Design

Based on the understanding reached by the evaluation of eXCeeD, we dedicate this section to describe how LeanCCD has responded to the study recommendations. In this proposal, the process can be summarized graphically in Fig. 2.

Fig. 2.
figure 2

Graphic summary from LeanCCD’s steps and activities

Full size image

  • Enabling the participants to carry out the activities:

Due to the problems of carrying out some eXCeeD activities as proposed, we invested in defining ways of presenting the process by creating guides that supported the application of each activity. We also made a presentation to the participants before the case study that evaluated the LeanCCD. Then, both participants and the researcher practiced some LeanCCD activities on a dummy project.

  • Include an initial step for the analysis of users and their tasks

Motivated by the lack of knowledge of the users’ needs indicated by the study that evaluated eXCeeD, we included a brief preliminary analysis of users and their tasks in an initial cycle. The activities proposed by eXCeeD were accommodated in a step called “Solution Detailing,” which is repeated in every cycle except the first, which happens in a step called “Defining an Initial Solution”.

The initial cycle (also called cycle 0) proposes to carry out a workshop, – with the participation of users, sponsors, developers and other stakeholders – contextual inquiry and the creation of an initial broad solution model, which represents a preliminary draft of the users’ goals and which answers the need for a comprehensive overall view of the solution, a deficiency pointed out by practitioners of agile methods [13]. The workshop is divided into three steps:

  1. 1.

    Product Vision Definition: activity inspired by the product vision statement, popular technique among practitioners of an agile method called Scrum [14]; it succinctly defines the product name, its class of product (e.g.: game, web application, website, portal, etc.), who should use it – with what goals and benefits and what the advantages over its competitor are. Unlike the vision statement, the target audience is divided between primary and secondary and a list of competitors and differentials can be articulated.

  2. 2.

    Target Audience Analysis: it is dedicated to expand the understanding of the primary and secondary target audience – defined in the previous activity. The activity makes use of empathy maps, a technique to promote co-creation, commitment, creativity, and innovation [15]. The goal is to define who these people that fit in that role are; what they do, say, hear, see, think, and feel. Through this reflection, it is possible to identify barriers and opportunities that can be addressed by the solution.

  3. 3.

    Definition of the Designer’s Deputy’s Profile: the goal is to make explicit the decisions on the communicative approach of a designer’s deputy to each role or to a combination of a role and a user profile. The designer’s deputy’s profile consists of the important utterances, attitudes, posture, tone, dialogue length and the communication channels that the user users to communicate with the system.

After the strategy workshop, the General Solution Modeling is initiated. During the first iteration, LeanCCD proposes that a simplified version of MoLIC represent the overview of the solution from the perspective of a user role. LeanCCD proposes to create a user goals diagram and an interaction model in a single representation [16], unlike the initial strategy of creating MoLIC diagrams [17, 18].

Creating less detailed – or even incomplete – diagrams in the early design stages promotes the designer’s reflection on alternative solutions. More elements are added as the interaction specification evolves [19].

At each solution detailing iteration, the designers refine the definitions made in the first stage. They also expand the general solution model, integrating the created interaction model fragments. Thus, the process defines a solution iteratively and incrementally. However, unlike eXCeeD, LeanCCD takes time to an Initial Solution Definition up-front, a common practice to design in agile software development [20].

  • Investigate the explicit usage of questions that support conducting interviews with users

The designers have not adopted the questions explicitly in the study that has evaluated eXCeeD. LeanCCD maintained the questions, integrating them in a guide and using sticky notes in bulletin boards as an incentive strategy of their use.

The questions were mostly used during the detailing of user goals. LeanCCD also makes use of the questions proposed by Aureliano in order to detail a set of user goals. However, we propose a User Target Board that groups the questions into categories according to the related topic. Figure 3 shows the questions in the guide for each area where the notes should be posted.

Fig. 3.
figure 3

User’s goals detailed guide with questions contained in the respective guide.

Full size image

  • Investigate the adoption of models and regular notes

The use of MoLIC models that were built in parallel with the user interface sketches is also part of LeanCCD. We kept this strategy as another opportunity to assess the adoption of MoLIC models in streamlined design approaches. We added to the regular notes the use of templates to record the information.

We also decided to propose the creation of conversation sketches, which are examples of conversations arranged in two lanes that represent, respectively, user utterances and system utterances, similar to essential use cases [21]. Instead of representing user intents and system responsibilities, conversation sketches are likely to represent a linearization of a path through a MoLIC interaction model.

The conversation sketches have two main objectives: to assist in the validation of the understanding with stakeholders that do not necessarily know the MoLIC language and to serve as rough sketches of interaction models, enabling the designer to explore alternatives through the simple manipulation of sticky notes.

4 Case Study

We evaluated LeanCCD happened in the software supply division of the Information Technology (IT) department of a multinational energy company. The division has a User Experience Center of Excellence founded with the goal of including HCI design techniques in the company’s software development process. The UX Center of Excellence is composed by professionals with graduate degrees in Design, Computing and Psychology.

The study aimed to answer the questions: “How is interaction designed when adopting LeanCCD?” and “How can LeanCCD’s evaluation collaborate to expand the understanding about communication-centered design processes?”

Fig. 4.
figure 4

Graphic summary of LeanCCD stages and activities

Full size image

The study lasted three days. The first day was devoted to the presentation of the process and a survey of previous experiences of the designers. In the other two days, the researcher observed the designers working on a project that aimed to develop a dashboard that presents metrics of call center tickets of IT solutions. Two stages of evaluation were held, the first one at the end of the strategy workshop and the second one at the end of the project.

As the study aimed to evaluate a process with few participants, we decided to perform a mainly qualitative research, but with some quantitative indicators that offer an overview of the opinions of the participants. The choice of a mixed methods research allows the collection of data through different instruments in order to triangulate the results and to make a more comprehensive discussion about the process.

The evaluation covered the following topics: clarity, usefulness, ease and intention of adopting each process activity. Throughout the assessment, the designers completed a questionnaire by technique, in which they assigned values between 1 (negative extreme) and 7 (positive extreme) for each topic. The researcher also explored the opinions in depth through individual interviews. At last, the evaluation included an evaluation activity in which each participant could assign to the guides of each technique one out of eight stickers representing three strength points, three weak points, one extremely strong point and one extremely weak point of the technique. In the final evaluation, the same amount of stickers was available to attach to a process graphic summary, as presented in Fig. 4.

5 Findings

The main LeanCCD evaluation findings are: the positive assessment of the initial cycle, the difficulty of assessing the general model in a small project, the abandonment of MoLIC models and the subsequent adoption of conversation sketches, besides the explicit use of the questions by combining boards and guides during the detailing of the user goals.

The designers positively evaluated the workshop conclusion, although it was not possible to notice a more active collaboration of users in completing the boards, as intended. Moreover, the definition of the designer’s deputy’s profile was negatively assessed according to its utility, ease and intent to adopt it.

Given the limited scope of this study, it seems reasonable that the participants did not feel the lack of a representation of the entire solution. However, in order to prioritize and refine a user goal, designers consulted the product vision board. The hypothesis is that the product vision would sum up the solution in a more comprehensive way for small scopes.

The abandonment of MoLIC models – similar to what happened to the study that evaluated eXCeeD – and the designers’ argument that the conversation draft is more aligned with a simplified approach raises a concern about the potential of the representation adopted in this study and the reasons and impacts behind the recurrent abandonment of the models.

The combination of boards and guides was an important step to investigate the explicit use of the questions in this study. However, we noticed that there is a small space for reflection within simplified approaches. In addition to that, the construction of the boards with A3 sheets was appointed as a limiting factor to the collaboration between designers and other participants. We suggest that, in future studies, they should be printed or drawn in larger dimensions, such as flipcharts and white boards.

6 Discussion

The findings presented in the previous section make us reflect on which factors influenced the behavior and opinions of the participants in this study. In this section, we will discuss the main factors highlighted, such as the over-theoretical approach adopted, the alignment of the proposal with simplified design processes, and the impacts of the strict sequencing of activities, which may have fostered the perception of an excessively methodical approach.

The perception of clarity on the activities of LeanCCD – expressed by the participating designers – was mainly harmed by the strategy of maintaining theoretical elements of the Semiotic Engineering and of MoLIC as proposed by their authors. Terms like “designer’s deputy” and “perlocutionary effect” were not understood or properly interpreted.

This overuse of theory creates issues to be analyzed in future studies: What is the knowledge needed to apply the process, and how can it be transmitted to an HCI design team without prior knowledge of SemEng or MoLIC? How can this knowledge be simplified and be adapted to the language of designers, facilitating their understanding and application?

Since the theory was conceived as academic research, the presentation of Semiotic Engineering concepts – if necessary – should follow a more technical discourse and be more compatible with their interlocutors, designers in the software industry. Successful examples of the technical discussions involving developers can be observed in the research of communicability evaluation methods [22].

Regarding the adoption of models, the analysis of the participants’ previous experiences pointed out prototyping as a common practice, but not combined with the construction of interaction models. As we could expect, designers have not adopted the interaction models. Once again, theory overuse may have influenced that decision. However, the participants of the study that have evaluated eXCeeD had an experience with the use of MoLIC and they also abandoned the use of the model. Would this repeated abandonment of MoLIC models on streamlined design processes indicate that the designers only see benefits in using this kind of models in projects with higher complexity?

Proponents of MoLIC pointed out that the main distinction is the representation of the entire system from the point of view of a user role. Although this is a gap of agile methods, the lack of such a representation has not been noticed in studies that have evaluated eXCeeD and LeanCCD.

It is possible to suppose that in streamlined contexts of design, the superficial view provided by the conversation sketches would be sufficient for designers. However, what would the designers lose by abandoning the modeling language or adopting less structured representations?

Further studies may investigate the effectiveness of the conversation sketches to represent the interaction and to analyze which elements from MoLIC would not be represented, and with what impact. The findings could point to the creation of a simplified version of MoLIC or to invest in expanding the use of conversation sketches in processes similar to eXCeeD.

During the final assessment using the process graphic summary, a designer scribbled and suggested changes to the order and combination of activities. This fact may highlight the need of flexibility to accommodate design preferences.

The sequencing of LeanCCD activity was more explicit during the evaluation process (Fig. 4) than in the graphic summary created subsequently (Fig. 2). That may have contributed to the perception of a methodical rigor in the way the process was introduced. Furthermore, the designers’ opinions on the use of some techniques, such as the proposal of the designer’s deputy profile, support the claim that some activities may have been executed only to fulfill a protocol.

7 Concluding Remarks

This article aims to contribute to an HCI design process that considers the communication between designers and users in the definition of a lightweight design process, the Lean Communication-Centered Design (LeanCCD). Among the main findings of this study were the positive assessment of the proposed initial cycle and the explicit use of the issues by combining boards and guides. However, we saw the non-adoption of MoLIC models – similar to what happened in the evaluation of eXCeeD – and we had difficulty in assessing the overall modeling of the solution due to the reduced scope of the projected system using the LeanCCD.

For future work, we recommend evaluating the general solution modeling in a study of broader scope – or in a less familiar area – and to conduct a few iterations to refine the model. A summary of the information obtained using the supporting questions is an interesting step, but it is necessary that some more comprehensive studies investigate their real use in agile contexts. Finally, they should invest in translating the boards and guides to other languages such as English and Spanish. This could expand the adoption of LeanCCD.

Among the factors that have most influenced the process execution are the theory overuse by the proponents of LeanCCD – indicating the investigation of a simplification of MoLIC and the concepts of SemEng. In addition to that, the rigid sequencing suggested by the graphic summary of the process may have encouraged the interpretation of an overly methodical approach, which goes against the principle of the studies that how designers usually behave [8]. By proposing a new graphic summary of the process activities, we attempted to avoid that impression (Fig. 2).

Giving up a methodical approach entirely implies to assume that we would always have experienced designers, working for companies able to deal with unpredictability, letting go of standardization, traceability and reproducibility, to list a few factors. We hypothesize that this is not a common scenario in which software development takes place in Brazil.

According to [7], “methods seem more like idealizations than prescriptions, and might better be presented as cases or exemplars rather than practical frameworks. This shift reveals the need to present a set of sound examples of how parts of various systems development method can be mixed and matched (perhaps with other, newly invented parts)”.

Thus, CCD, eXCeeD, and LeanCCD can be considered examples of how you can adopt a communication-centered approach to the design of interactive systems. Studies such as the one reported in this paper would help to choose the approach to follow, to increase the understanding of the adoption of communication-centered processes.

Therefore, we suggest the interpretation of methodical and non-methodical approaches as extremes of a continuum, which can encourage us to find the balance that motivated the creation of LeanCCD. This balance point should be able to assist inexperienced designers to find design strategies, without losing the flexibility to accommodate individual preferences and to react to situations related to the environment in which these processes take place.