Keywords

1 Introduction

The software development landscape continuously evolves, and agile methodology has delegated teams to adapt and deliver value in the dynamic work environment. Agile retrospectives are a capstone of the agile framework and a crucial practice to many software development teams [1]. The human element must be noticed in the agile retrospective cycle as it directly affects the success of the software development cycle [2]. Establishing a psychologically safe work environment is crucial for leading to positive and practical agile retrospective sessions. When the team reflects on the experience, areas needing improvement, and action plans at the end of each iteration, they express themselves by sharing thoughts and opinions [3]. While doing the same, psychological safety elements, i.e. emotions, are involved during the online retrospective meeting [4]. A couple of words expressed during the online meeting could lead to a negative or positive work environment [5]. Capturing emotions could be fruitful as it helps to detect harmful or favourable online behaviours [6], hinder or facilitate the software development cycle, moralize or demoralize the team, and nourish or discourage innovation and cooperation inside the organization. Hence, it is essential to gather the emotions of the agile retrospective teams [4]. Emotions are interconnected and related concepts placed under the roof of human psychology and communication. Emotions can be a range of feelings, for example, happy, sad, anger, fear, etc. [11]. The structure of emotions is the basis for creating human sentiment [7]. So, sentiment is considered the high-level category of emotions, categorized into three categories: positive, neutral, and negative. Over time, emotions are connected with certain experiences and beliefs that generate a sentiment [8]. This study revolves around emotions during the agile retrospectives of two software development teams. We aim to investigate the various emotions contributing to agile teams. Our research question is: Rq.) How often are different emotions repeated during the online agile retrospective? We conducted multiple case studies to detect the type of emotions and their frequency in the retrospectives from two software development teams. We found that several emotions, such as (approval, realization, excitement, relief, disappointment, etc.) overlap in both the agile retrospectives. Approval was repeated maximum (17 times) whereas pride, fear, embarrassment was minimum as it occurred only once.

2 Background and Related Work

2.1 Emotions in Online Agile Retrospectives

Software teams at the workplace express many emotions that impact their productivity. Girardi et al. investigate the correlation between developers’ emotions and productivity. The authors experimented with 21 developers from five Dutch software companies [9]. The study identified a positive correlation between developers’ emotions and perceived productivity. In addition, Graziotin et al. examine the effect of emotions experienced by software developers [10]. Based on the survey results from 317 participants, the authors found that emotions have some impact related to the happiness and unhappiness of developers. These developers practising retrospectives should feel psychologically safe [3], which encourages them to share their experiences and emotions [4]. A recent study by Grassi et al. describes the importance of emotions in agile retrospectives and how students’ emotions vary through performing activities in a software engineering course. The authors developed an emotion visualization tool that visualizes emotions, actions, and bio-metrics. Agile retrospectives were chosen as a test bed to evaluate the tool. The study shows that detecting emotions can assist in discussing and fixing various issues that arise in a sprint [4]. However, there needs to be more research that applies emotion analysis in online agile retrospective meetings. Often, it is noticed that retrospective participants use emojis to express emotions at various stages of the meeting, for example, during a chat [3].

Fig. 1.
figure 1

Theoretical Framework

Full size image

2.2 Theoretical Framework

This section specifies various emotions that we collected from literature [11,12,13,14], which serve as the base of the theoretical framework. Neutral emotions: Neutral emotional states that are neither positive nor negative. The literature outlines three neutral emotions and their following example. 1.) Confusion: “Ok, just making sure I was confused”, 2.) Curiosity: “I am curious to know about [something]”, 3.) Realization: “I figured/realized something” [11, 12]. Positive emotions: These emotional states or reactions are welcoming, nice, inspiring, and delightful. The literature outlines eleven positive emotions and their following example. 1.) Admiration: “Please keep up the great work”, 2.) Amusement: “Haha, actually, grandpa did! Go figure”, 3.) Approval: “We have received approval from the boss”, 4.) Caring: “He was caring for his dog”, 5.) Desire: “I can’t wait to hear the stories”, 6.) Excitement (Gratitude, Joy, Enthusiasm): “Excellent idea, thank you”, 7.) Love (Affection, Adoration, Cuteness): “Cause you were so tiny and fragile”, 8.) Optimism: “I am confident about it”, 9.) Pride: “We are the best”, 10.) Relief: “Thank god, I was just thinking to do it”, 11.) Surprise: “Wow, what a sunny day” [13, 14]. Negative emotions: Negative emotions are reactions that are unwelcoming, unpleasant, upsetting, and uneasy. The literature outlines eleven negative emotions and their following example. 1.) Anger: “If this is who you are”, 2.) Annoyance: “But the man keeps it tearing apart”, 3.) Disapproval: “She is not ready yet”, 4.) Disappointment: “vmware fusion seems to get slower and slower”, 5.) Disgust: “Well that made me want to continue to live in Alberta”, 6.) Embarrassment: “I feel foolish”, 7.) Fear (Anxiety, Nervousness): “Is Someone there”, 8.) Grief (Pain, Tiredness): “It’s back, what I mean is my headache”, 9.) Remorse (Guilt): “I am sorry, I wasn’t perfect”, 10.) Sadness (Distress): “Poor guy”, 11.) Surprise: “What, you won’t be two blocks away anymore?” [11, 14]. As shown in Fig. 1, we captured similar examples of emotions and mapped them with the (audio, text and icon) involved in agile retrospective meetings. With the help of the framework, we retrieved a list of emotions and their frequency presented in the agile retrospective.

3 The Study Research Process

We conducted multiple case studies to collect emotions from two software development teams. We selected the cases based on the convenience sampling approach [17]. The first case is a team (T1), a software company based in Germany that helps to calculate assessment management for many individuals and organizations. The second team (T2) is a multinational software company with several European offices. Data collection - We collected the data from three online retrospective meeting videos. (M). One video from (T1) and two videos from (T2). The team (T1) meeting (T1-M1: lasted around 35 min with 3 participants and used Trello and Zoom as software tools for OAR). From the two videos of T2, we used the first one as our pilot study (T2-M1: lasted around 15 min with 10 participants) and the other one as our case (T2-M2: lasted approximately 30 min with 10 participants and used a digital board called Parabol, Microsoft Teams). We converted the videos into text through cockatoo (software that converts video to text files). The text was used as our transcripts for analysis.

Table 1. Data Analysis
Full size table

Data Analysis. We applied the research approach called the “bracketing technique” to analyse the three videos [15]. This technique helps to describe precise time-stamped breakpoints and use them for coding. First, we analysed the pilot study (T2-M1), and later, we completed the analysis of T1-M1 and T2-M2 meetings. To analyse each time-stamp or chunk (1 min long), the authors manually listened to the audio first and then validated the text with the theoretical framework. Both authors together picked each minute chunk (few examples are visible in Table 1) one by one (chunks 1,2, and so on), assigned emotion labels based on the theoretical framework, and reached a consensus on the identified emotion. Although the retrospectives lasted for around 30–35 minutes, we found only 28 min of instances or chunks for T1 and 17 chunks for T2 due to the following reasons. We excluded chunks were: 1.) Teams had no audio content relevant to the retrospective available that could be converted to text; 2.) The team reflected or thought during the period; hence, no conversation or text was shared during the meeting. After analysing the manually collected emotions of the chunks, we first used software tool Text2data and then ChatGPT-3.5 to analyse the text and validate our results. We found out that our study was similar to ChatGPT analysis compared to the tool. We discovered that the tool only used minimal emotions to calculate compared to our manual calculations based on theoretical framework. The software tool considered only fifteen types of emotions: “anger, boredom, emptiness, enthusiasm, fear, fun, happiness, hate, joy, love, neutral, relief, sadness, surprise, and worry”. In contrast, the theoretical framework 2.2 in the previous section consists of twenty-five types of emotions. We also asked ChatGPT what methods and algorithms generated the analysis. The ChatGPT used the following (satisfaction, happiness, agreement, contentment, joy, approval, optimism, dissatisfaction, concerns, criticisms, disappointment, frustration, scepticism) emotions for the analysis.

4 Findings

Fig. 2.
figure 2

Name of the emotions and their repetition in T1-M1 retrospective

Full size image
Fig. 3.
figure 3

Name of the emotions and their repetition in T2-M2 retrospective

Full size image

Figures 2 and 3 present the type of emotions in the online retrospective meetings. In both figures, the X-axis represents the number of (times) or frequencies the emotions were repeated, and the Y-axis represents the type (name) of the emotion.

Neutral emotions: We can observe that Realization (9 times repeated) and Curiosity (2 times repeated) were the two common neutral emotions in retrospectives. It shows that retrospective members were either realizing or curious about the sprint’s past, present, or future tasks. For example, a participant realized: “we didn’t probably think it through completely. We ended up completing it. But probably in the other direction, so now we have to consider it and the next steps to see how we can go back on our steps. Let’s go on with one of the other cards.” (T1-M1). Whereas another team member was curious: “OK. Can we put the [task] inside the sprint?”, “So maybe we can start with the collaboration and coordination between the two teams if you agree?” and realized: “OK, got it, so probably we should discuss the two deltas and this one here if you want.” (T2-M2).

Positive emotions: Observing the Figs. 2 and 3 positive emotions, Approval (17 times repeated) was the most preferred whereas the second most was Excitement and Relief from T1 and Admiration from T2 retrospective. Regarding Approval, one participant mentioned: “Yeah, totally agree with that. I mean, we are both doing.” “We could be good at this one for an action point, right? What do you say? Yeah. What could we do this, actually." (T1-M1). The same team was also excited and relieved: “I’m delighted. Three people have already said yes. So, I’m pretty happy with it. Yeah. And this can lead to a lot better estimates.” (T1-M1). The second team encountered an admiration moment where the participant quoted “Thank you to [Names] for your continuous patience and help during this sprint. [Names], best teammates ever, and thanks to have followed the DB activity.”(T2-M2).

Negative emotions: Concerning the negative emotions, both teams had a Disappointed (on team 7 times repeated) feeling. Second repeated, Embarrassment or Fear for T1 and Disapproval for T2 as a negative emotion during the retrospective. The team was disappointed and quoted “Delta, team A, and B working on the same project, with no coordination at all. Delta, Are eight story points issues too big? How can we avoid the failure of the sprint?” (T2-M2). The team T1 had a fear about estimation as they mentioned “So, let’s just be careful. Yeah, it is affecting too much the planning for Q1 for those estimations?” (T1-M1) Whereas there was a moment of disapproval as one member mentioned “No, I disagree with this. So this was not the idea of the teams, I think. No. The teams should be independent.” (T2-M2).

5 Discussion

Our study sheds light on various emotions during the online agile retrospective. Emotions are intrinsic to human communications, and our findings suggest that they can help retrospective groups shape better outcomes and learnings. Within the two software case study teams, we found in total eighteen emotions (see Figs. 2 and 3), namely [3 neutral (Realization, Confuse, Curiosity), nine positive (Approval, Excitement, Relief, Optimism, Amusements, Admiration, Desire, Pride, Gratitude), and six negative (Disappointed, Disapproval, Sadness, Annoyed, Embarrassment, Fear)] emotions. We also identified the overlap of various emotions between the two cases such as (Realization, Curiosity, Approval, Admiration, Pride, and Disappointed). Knowing the emotions can help to encourage psychological safety, strengthen empathy, and generate pain points and insight in a team. But to grasp the emotions, the team must respect confidentiality and treat all the members with respect in the company. We observed that factors like the company’s culture and the scrum leader’s behaviour facilitating the retrospective could influence emotions. Moreover, a tone could also affect the comfort level of participants and change of mindset to discuss the task positively or negatively. This study lays direct implications for agile practitioners. Retrospective teams can create an environment to encourage communications with open expression of positive emotions and constructively managing negative emotions. Teams could focus better on the improvements of a cycle and apply some methods to solve the negatively evoked issues before the end of the retrospective. A team could use tools, as mentioned in the study [4], that could capture emotions during retrospective sessions. Concerning the limitation of this study. It was conducted with only three retrospective videos. We had a limited number of videos because retrospectives are a practice that occurs at the end of the sprint cycle [16], but usually, it is longer than other meetings. Hence, we selected an agile retrospective for the study. This limits the generalizability of our findings. Future research could involve additional sessions of retrospectives, sprint planning, daily planning, daily stand-up, and product feedback that could lead to a better understanding of both sentiments and emotions in online agile retrospectives.

6 Conclusion

Human emotions are the factors that affect the success of agile retrospectives. In this paper, we study the emotions in online agile retrospectives from two software teams by identifying how often emotions are repeated throughout the agile retrospective. Our study reveals that approval, excitement, admiration, and relief are the most positive emotions. Disappointment and Disapproval are the most frequent negative emotions. At the same time, realisation and curiosity account for neutral emotions. Emotions are crucial in shaping the digital interaction, team dynamics and decision-making process. Revealed emotions act as a facilitator that affects the performance of a team. It is vital to foster the trend of psychological safety in agile retrospectives so that teams in organizations can boldly express their emotions, leading to improved sprint cycles. In the future, the additional research should encompass sentiments obtained from emotions, which could further enhance the entire software development process.