Abstract
This integrative literature review aims to review, critique, and synthesize the existing literature on appreciative inquiry (AI) in three disciplines where the most significant number of AI-related articles could be found: healthcare, higher education, and management. We sought to identify critical insights, along with similarities and divergencies between the three fields. We identified diverse reasons for adopting AI methods and differing degrees to which researchers adhere to the 4D cycle. The reviewed literature suggests that AI yields positive effects across three levels (individual, group, and organization). Our results indicate that positive impacts from AI can be achieved even when not all steps of the 4D cycle are strictly followed. Furthermore, we discovered numerous innovative AI applications and multiple instances of using modified AI models. Based on our research findings, we proposed a systems model for understanding the AI process and offered implications for both HRD research and practice.
Appreciative Inquiry (AI) represents a positive approach to driving organizational change (Drew & Wallis, 2014). The AI model is built upon the work of Cooperrider and Srivastva (1987), who questioned the general belief that organizational change interventions should be problem-based (Knibbs et al., 2012). Whereas a traditional problem-solving approach assumes the need to identify problems within the organization and focus on providing solutions to these problems, AI begins by appreciating and valuing the assets and strengths of the organization (Cooperrider & Whitney, 2005). It assumes that there are at least some elements that function effectively within every organization, and valuable insights can be gained by asking about those positive aspects of the organization (Knibbs et al., 2012).
Since its introduction, AI has been a catalyst for revolutionizing the field of organization development (OD), leading to significant improvements in organizational performance across people and profit (Bhattachary & Chakraborty, 2020). AI has played a substantial role in enhancing efficiency and performance, especially within North American workplaces, driving transformative changes in large organizations such as NASA and McDonald’s (Watkins et al., 2016). Although most of the early applications of AI can be found in management and OD work, over the years substantive streams of AI application have emerged in many other disciplines and areas of practice, especially in healthcare, including nursing, pharmacy, and healthcare education, and in higher and K-12 education.
In the past decade, several attempts have been made to systematize the knowledge about the AI approach through literature reviews. Watkins et al.’s (2016) integrative literature review sought to identify the impact of AI on changing clinical nursing practice in in-patient settings. Merriel et al. (2022) aimed to examine AI’s effect on healthcare through systematic review and narrative synthesis. However, those attempts were limited to reviewing a comparatively small number of healthcare literature from biomedical databases, such as Medline and Embase, and the focus has been on relatively small subsets of the healthcare field. We could not identify any review articles aimed at understanding AI from a broader perspective, integrating the literature from more than one area, and incorporating fields other than healthcare. The absence of such studies is especially noticeable in HRD since AI has become a widely used OD approach. We could find only a small number of studies researching AI practices, or literature reviews, related to this approach, published in HRD journals. Indeed, we were unable to find English-language peer-reviewed articles featuring “appreciative inquiry” in their titles within the HRD/HRM journals over the past decade. Given the scarcity of AI-related research in HRD and the much greater prevalence of this research in several other fields, it would be beneficial to gain insights from related fields that could inform HRD practice and future research.
To address the gap described above, we aimed to review, critique, and synthesize the existing literature on AI in healthcare, higher education, and management. We selected these three disciplines because our preliminary scoping review determined that the largest number of AI-related articles can be found in these fields. We focused on empirical research, since a significant body of such research emerged in each of the three fields, and, therefore, synthesis of such research could produce substantive conclusions, applicable to HRD and other related fields. The research questions are as follows.
What are the key insights from the AI-related literature?
What are the differences and similarities among the three disciplines?
The integrative literature review method was adopted to identify key insights from each of the three fields and to explore similarities and/or divergencies between fields. Furthermore, the goal was to use the insights gained from the review of the literature from these three fields to develop recommendations for advancing AI-related research and practice in HRD.
We start by providing a concise overview of the AI method and a review of the past and current research on AI published by key scholars and practitioners working in traditional spheres of application of AI in organizational change and development. We then describe our method, including the literature search and analysis procedure. Next, we summarize our findings and describe the themes that emerged in our analysis, highlighting differences and similarities among the three areas: healthcare, higher education, and management. While the first author does not have experience with practical applications of AI, the second author has experience in consulting work associated with AI and teaches AI concepts in graduate classes, which may have led him to have biases about AI. Recognizing this, we cross-validated our interpretations of the emerging themes, striving to maintain overall neutrality. Finally, we discuss implications for HRD research and practice.
Appreciative Inquiry: An Overview
AI is a collaborative process that explores, identifies, and enhances the best of “what is” in organizations intending to create a better future (Preskill & Catsambas, 2006). The initial conception of AI dates back to 1980 when David Cooperrider, a doctoral student, and his thesis supervisor, Suresh Srivastva, first formulated the concept while working on an organizational development project (Watkins & Mohr, 2001). They suggested AI as an alternative to traditional action research (AR), trying to address AR’s limitations, mainly associated with its heavy reliance on a problem-centered approach (Fry et al., 2002). Whereas the problem-solving approach assumes an organization is “a problem to be solved”, AI considers it as “a mystery to be embraced” (Cooperrider & Whitney, 2005, p. 13).
There are eight principles of AI: (1) The constructionist principle, (2) the simultaneity principle, (3) the poetic principle, (4) the anticipatory principle, (5) the positive principle, (6) the wholeness principle, (7) the enactment principle, (8) the free-choice principle (Whitney & Trosten-Bloom, 2010, p. 52). These eight principles primarily center around the impact of positive imagery and mindset and highlight the significance of conversations in AI (Preskill & Catsambas, 2006). Incorporating these principles, AI emphasizes the importance of social interaction in forming knowledge and the energy generated by individuals experiencing positive emotions such as hope, empathy, connection, joy, and inspiration.
The traditional 4D cycle of AI refers to the four phases of the AI process: discovery, dream, design, and destiny (Figure 1). According to Cooperrider and Whitney (2005), the initial discovery phase involves engaging all stakeholders to identify and articulate strengths and best practices within the organization. In the following dream phase, a transparent results-oriented vision concerning discovered potential is created. In the design phase, the organization’s structure and processes are re-designed to amplify the positive core and bring the newly formulated dream to fruition. Lastly, the destiny phase aims to strengthen the affirmative capabilities of the system, ensuring sustained momentum for long-term change and performance. Appreciative inquiry 4D cycle (adapted from Cooperrider et al. (2008)).
Whole-system inquiry and AI summits are the two most popular and effective methods using the 4D cycle (Cooperrider & Whitney, 2005). The whole-system inquiry is an AI method that involves all organizational stakeholders (Whitney & Trosten-Bloom, 2010). The AI summit can also be characterized by its large-scale process involving, ideally, all the stakeholders, but it differs from the whole-system inquiry in that the AI summit consists of four days (one day per each of the stages of the 4D cycle) (Ludema et al., 2003). The AI summit is reported to maintain its benefits even when held in virtual settings (Conkright, 2011).
Bushe and Kassam (2005) described cases where AI resulted in transformational changes. They reported two key elements of AI that distinguish it from traditional approaches to OD and change management: (1) a focus on changing individuals’ mindset rather than their actions, and (2) a focus on self-directed change processes that emerge from new ideas.
AI is associated with the field of positive psychology at its core, as it was created as a response to the deficit approach prevalent in both the research and practice of psychology (Seligman & Csikszentmihalyi, 2000). According to positive psychologist Frederickson (2001, 2013), positive emotions can enhance human flourishing, facilitate resilience in overcoming challenges, foster meaningful relationships, and drive sustained behavioral changes. Positive psychology offers scientific evidence supporting AI usage and the strengths-based movement within organizations (Wocken, 2020).
Studies in leadership and OD report advances in AI. Hart et al. (2008) found that AI helps support individual leader development. Whitney and her colleagues (2010) suggested the concept of appreciative leadership. In their book, appreciative leadership was characterized by its capacity to unleash one’s creative potential and transform it into positive power, ultimately impacting the world positively. Sloan and Canine (2007) applied the philosophy and practice of AI to coaching and referred to the process as appreciative coaching. There was also an attempt to apply AI philosophy to strategic planning. As an alternative to the widely-adopted SWOT approach, Stavros and Hinrichs (2009) suggested the SOAR approach, which focuses on strengths, opportunities, aspirations, and results.
Critics argue that AI, in its exclusive focus on positive narratives, overlooks the potential for positive changes that may emerge from seemingly negative experiences, including situations like embarrassment, anger, anxiety, fear, or shame (Barge & Oliver, 2003; Fineman, 2003). If the motivation behind “emphasizing the positive” is to evade addressing genuine concerns or to stifle dissent, AI has the potential to transform into a disguised form of repression (Fineman, 2006). Moreover, AI methods could potentially conceal disagreements, particularly if discussions take place in contexts where power inequalities exist (Aldred, 2011).
Method
An integrative literature review is “a form of research that reviews, critiques, and synthesizes representative literature on a topic in an integrated way such that new frameworks and perspectives on the topic are generated” (Torraco, 2005, p. 356). It allows researchers to provide a concise overview, analysis, and integration of existing literature within a particular academic field (Chermack & Passmore, 2005).
Torraco (2016) emphasized that one of the distinctive features of integrative literature reviews is a detailed description of replicable and systematic literature searches. Further, Torraco explained that an article, reporting the results of an integrative literature review, should explain the progression of the study’s synthesis and discoveries and illustrate how the main themes were derived from the existing literature. According to Callahan (2010), a well-written integrative literature review-based article articulates the following: (a) databases and search engines used, (b) timeframe of the literature search, (c) the person who searched, (d) keyword combinations for the search, (e) number of articles retrieved (for each keyword and the total), and (f) inclusion criteria for the review (p. 301). In our study, we used the above points, suggested by Callahan, as well as the detailed guidelines for integrative literature reviews, formulated by Torraco (2005, 2016).
The data collection started by searching two databases: EBSCOhost EJS and Web of Science. These databases were selected since they cover most social science publications, available through the library of the major public university where both researchers work. In addition, to make sure that our preliminary scoping review did not miss any HRD-based empirical articles on AI, we conducted a targeted search in HRD and human resource management (HRM) journals. The following journals were included: Human Resource Development Review (HRDR), Human Resource Development Quarterly (HRDQ), Advances in Developing Human Resources (ADHR), Human Resource Development International (HRDI), European Journal of Training and Development (EJTD), Human Resource Management, Human Resource Management Review, Human Resource Management Journal, and International Journal of Human Resource Management. The initial literature search was conducted in December 2022. We included the articles that were published from 2012 to 2022 to ensure that we capture the recent trends and discussions. The major keyword used for the literature search was “appreciative inquiry” as there is no other synonymous expression. Our search criteria required that the articles must have “appreciative inquiry” in their title. Only English-language, peer-reviewed articles that provide full access to the manuscript were included in the initial search.
The search resulted in a total of 236 articles. In HRD/HRM journals, there were no articles about AI adhering to the stated criteria. After removing 43 duplicates, 193 articles were included in our database. We manually classified the areas of scholarship by looking at the article titles and journal titles of searched items. Among 193 articles, 78 were healthcare-related, 24 - management-related, and 19 were related to higher education. The management category included articles in sub-fields or closely related fields, like leadership and organization studies. The rest of the articles were in numerous other fields, with just a handful of articles in each of them. During the screening process, we reviewed each article’s title, abstract, and keywords and excluded articles based on the following criteria: (a) the articles that are not in proper format (e.g., no abstract), (b) the articles that are not peer-reviewed (e.g., book review, essay, or interview). After the screening process, 105 pieces were left: 70 healthcare, 17 management, and 18 higher education articles. Among healthcare articles, 13 (18.6%) were education and training related.
The remaining 105 articles were subjected to in-depth analysis. For each article, we reviewed the full manuscript and recorded several key points in an Excel spreadsheet: research focus, characteristics of studies, research participants/subjects, purpose, methodology, and results. In this process, we were capturing emerging themes and actively took notes. The analysis started with management articles and continued with higher education and healthcare articles. After the first round of analysis, we revisited each of the three scholarship areas to check whether the themes found in the later analysis were also present in the previously reviewed materials. In addition to finding commonalities in all three areas, the analysis was focused on discovering the uniqueness in each area. For the four conceptual papers and four literature reviews, which were only included in the final pool to comprehensively report the methods used for AI studies, we did not review the entire manuscript nor record their key points, leaving them out from the rest of the analysis.
Findings
In this section, we report the seven themes that emerged from the analysis: methods used in AI studies, purposes of using AI, positive effects of AI, innovative applications of AI, usage of modified or alternative models, acknowledging and overcoming inherent limitations of AI, and the efforts to further develop the AI model.
Methods Used in AI Studies
Research Methods Used in the Reviewed Articles.
Other methods were seldom adopted in AI research. For example, there were only four quantitative studies: three in healthcare and one in management. The experimental design was used in three of the reviewed papers (2.9%). Scala and Costa (2014) studied two groups: one received AI coaching and the other served as a control group. In Mandal’s (2022) study, groups were involved in team-building and AI training interventions in a controlled experimental setting. The effects of the interventions were evaluated by measuring the dependent variables, which included group communication and group decision-making.
In healthcare, a longitudinal cohort study (Hussein et al., 2014) was utilized. And one of the healthcare studies reported the results of the development of a survey instrument (Netthong et al., 2020).
Compared to management and healthcare, most articles from higher education did not provide clear and well-documented explanations of the utilized methodology. Often, it was necessary to infer the methodology from subtle indications within the study. More specifically, it was difficult to tell whether a study is a general qualitative study, case study, or based on action research. For example, in one instance (Fileborn et al., 2022), one could argue that the authors used case study approach because the project was focused on one bounded case, but no description of a specific case study method, utilized by the authors, was provided.
Purposes of Using AI
There were various purposes for using AI in projects, described in the reviewed studies, and the degree to which the 4D cycle was implemented accounted for some of these differences. We found three major purposes for the use of AI. First, AI was used to achieve desired changes in organizations. The type of possible changes facilitated by AI included cultural (Halm & Crusoe, 2018), behavioral (Scala & Costa, 2014), and organizational (Hussein et al., 2014) changes. Projects that employed all four phases of the 4D cycle belong to this category. For example, in Hung et al. (2018), AI was used to facilitate collaboration among practitioners to develop knowledge and drive action for change. Such use of the entire 4D cycle was a popular approach in all three areas, but, percentage-wise, the management field was most invested in this approach.
Second, some researchers used AI to uncover the object of appreciation in past positive experiences, using only the first step (e.g., Bushe & Paranjpey, 2015; Chodzaza et al., 2020; Schmied et al., 2019) or the first two steps of AI (e.g., Butcher et al., 2022; Mdoe et al., 2021; Moore et al., 2017). This was a trend often found in healthcare and higher education. The overarching question was: what was positive about the previous experience? For example, Helms et al. (2012) sought best practices and asked participants to nominate the best three exemplars. In some cases, there were additional reasons for partially incorporating the 4D cycle in a study, such as time constraints (Naaldenberg et al., 2015) and concerns about burdening the participants (Moore et al., 2017).
Lastly, AI was adopted as a research method in some of the studies. For example, He and Oxendine (2019) collected data through interviews and used AI only as a lens for thematic analysis. Calabrese (2012) described AI as a theoretical research perspective and utilized it to assist participants in discovering inherent leadership strengths and successful experiences. This approach was widely observed in higher education and to a lesser extent, in healthcare.
Some of the articles that claimed using AI as a research method remained unclear about how specifically AI was applied (e.g., Frerich & Murphy-Nugen, 2019; Hung et al., 2018; Jongudomkarn & Macduff, 2015). For instance, Frerich and Murphy-Nugen (2019) did not explain how they used the AI method in their study and only mentioned that AI provided an “emancipatory and empowerment perspective” (p. 12).
Positive Effects of AI
The majority of the reviewed articles addressed the positive effects of AI at one of the three levels: individual, group, and organization. The reported positive effects were most prominent on the individual level. After participating in AI-driven projects, individuals showed strengthened morale (Allen & Innes, 2013) and an enhanced sense of belonging and collaboration (He & Oxendine, 2019). Wall et al. (2017) reported managers having more positive emotions and making stronger change efforts after using AI as part of their projects. Verleysen et al. (2015) found that people with direct experience with AI felt that this experience was energizing and generative. Trinh et al. (2021) reported that AI facilitated empathy and evoked a sense of larger purpose and a sense of belonging to a group which, in turn, added meaning and value to individuals’ lives.
In addition, according to some of the reviewed studies, AI offers opportunities for individuals to uncover new possibilities and boost performance. Studies reported increased social capital (Calabrese, 2012), increased expertise in the program contents (Jewell & Murphy, 2020), positive performance changes (Johnson, 2014), increased desirable health-related behaviors (Olayinka et al., 2020; Savage et al., 2018), and improved well-being (Lane et al., 2018). Notably, Diedricks et al. (2018) revealed that students with HIV recognized their strengths and discovered the untapped potential for living with HIV through their participation in AI. Fernald et al. (2020) reported improvement in participants’ attitudes toward mobility treatment after participating in AI workshops. Watson (2013a) emphasized that AI stimulated a high level of engagement among managers and expedited the fostering of relationships, motivation, and innovative problem-solving.
Among studies reporting group-level effects of AI, the largest number demonstrated how participating in AI allowed groups to obtain a collaborative spirit and strengthen their feeling of unity. AI helped bring the practitioners together to take specific steps in changing their practices (Fowler-Davis et al., 2022) and allowed a unique opportunity to establish positive connections and relationships, which led to transformational change within the group (Oxendine et al., 2022). Moreover, Mandal (2022) showed that participation in AI led to enhanced group communication, decision-making, membership satisfaction, and performance. According to Halm and Crusoe (2018), as the result of participating in an AI session, the team was able to develop a culture of clinical excellence in the organization. Other studies emphasized the benefits of using AI in building rapid relational trust and bonding between individuals, which may be particularly advantageous in fostering group cohesion (Ng et al., 2019; Sharma et al., 2015).
Finally, some of the reviewed studies demonstrated how AI might provide benefits at the organizational level. AI facilitated organizational change (Hussein et al., 2014), helped to establish a shared vision, identify strategies, and develop plans for continued collaborations and partnerships (Trajkovski et al., 2015). Heiser and Swallow (2015) reported that AI helped the company to plan key strategic initiatives based on its core strengths. The study authors also emphasized that the AI process created a high-energy environment that nurtured positive communication, collaboration, and enhanced engagement. Meanwhile, Watson (2013a) highlighted AI’s role in stimulating a collaborative mindset within the organization and providing an evaluation framework. It was argued that the AI process was generative to other institutional processes and helped the establishment of generative leadership (He & Oxendine, 2019).
Innovative Applications of AI
Whereas AI was initially suggested as an organization development intervention by Cooperrider and Srivastva (1987), some of the reviewed articles reported efforts to apply AI differently from its original purposes. Higher education researchers were the most proactive in expanding the application of AI. One of the important areas of such expansion was applying AI to program and instructor reviews. AI framework was used for practical course evaluation (Allen & Innes, 2013; Kung et al., 2013; Sacco-Bene, 2022). Sacco-Bene (2022) mentioned that while AI offers a substitute for traditional course evaluation, AI itself is more of an approach with a set of strategies than an evaluation method. Oxendine et al. (2022) also described how to use AI in peer evaluation using AI and suggested Appreciative Peer Evaluation Meeting Guide (APEMG). The guide utilizes a general AI framework and the 4D cycle to foster positive discussions and improve performance appraisals. By focusing on what is working well and encouraging reflection on past successes, individual strengths, and opportunities for professional development, the APEMG aims to enhance collaboration among faculty members and facilitate transformative goal setting. The process includes feedforward to provide constructive feedback and engages tenure-track faculty members in self-evaluation.
Other innovations in using AI were related to teaching methods. According to Calabrese (2012), the undergraduate students were invited to reflect and respond to AI blog posts which stimulated reflective responses emphasizing their leadership competencies and experiences. Jewell and Murphy (2020) used the AI framework during the onboarding process at a university library scholar program. Johnson (2014) discussed the concept of “appreciative andragogy” and explored the application of AI as an online instructional strategy with nine online class facilitators.
Studies reporting innovative use of AI in the healthcare field were mostly education and training-related. Similar to findings in higher education, multiple studies reported using AI as a teaching strategy (Butani, 2019; Chauke et al., 2015; Dewar & MacBride, 2017). For example, Dewar and MacBride (2017) employed AI as an effective educational strategy to foster positive human interaction. They emphasized AI’s role in promoting the engagement of individuals instead of regarding them as mere project beneficiaries. Meanwhile, Ghosh et al. (2022) used modified AI-guided focus group discussions and named the process SCORE (Strengths, Challenges, Opportunities, Results, and Evaluation).
In the field of management, there were only two articles reporting innovative uses of AI. Watson (2013b) discussed AI’s application as a diagnosis and evaluation tool, reporting that AI had promoted maximum engagement with learning and development within a group of managers. And, Trinh et al. (2021) successfully incorporated the AI method in a large MBA course to foster a hospitable learning space.
Usage of Modified or Alternative Models
Different/Modified Models of AI.
The traditional 4D cycle is composed of four phases: discover, dream, design, and destiny (Cooperrider et al., 2008). The last stage is sometimes labeled as the “deliver” stage (Dempster & Kluver, 2019; He & Oxendine, 2019), which further emphasizes the importance of implementing the action plan. Meanwhile, Dewar et al. (2016) suggested the AI process consists of four phases with different labels. Those steps were utilized in Dewar and MacBride (2017) and McSherry et al. (2018). According to Dewar and MacBride (2017), a specific question is asked at each of the stages: “discovery”—what is working well, “envision”—what would you like to see happening more of the time, “co-create”—what do we have to do to achieve our vision, and “embed”—what has worked well and how people can be supported to develop further (p. 1378). The suggested labels by Dewar et al. are more descriptive of the activities involved in each stage, although the content of each step remains essentially the same as in the traditional 4D cycle.
Tripathi et al. (2020) suggested another modification of the AI cycle, calling it the 4I-cycle. The 4I cycle consists of inquire, imagine, innovate, and implement steps. One unique aspect of this cycle is that the barriers and enablers are recognized in the first “inquire” phase. Then, in the following “imagine” step, the participants look for ways to affect the barriers and amplify the existing enablers. Based on that analysis, they clarified the affected stakeholders and set implementation plans in the “innovate” stage.
Some studies introduce another step that precedes the main four steps. Guo et al. (2022) used the “inception” phase, where they “formed an intervention group, introduced the aims and contents of the program, and constructed affirmative themes” (p. 7). Similarly, Dewey et al. (2022) used a “define” phase, where the following question was asked: “What is the subject of inquiry?” (p. 1399). McSherry et al. (2018) did not articulate what specific tasks are conducted in the “setting the scene” stage, but it can be assumed that the preparation for the AI is done in this phase.
In addition, follow-up steps were suggested in several articles. For instance, Guo et al. (2022) proposed a follow-up step named “keep.” With the “keep” phase, authors aimed to “strengthen personal and organizational resources” and “continue setting goals and work plans” (p. 7). Researchers established a support team to enhance the work performance of participants during the destiny phase. In the subsequent keep phase, they focused on strengthening personal and organizational resources through performance feedback. As such, the action plan developed through the AI was further reinforced in the follow-up step.
Acknowledging and Overcoming Inherent Limitations of AI
AI is inherently focused on finding the positive and successful aspects of practice. Thus, it is easy to ignore negative practices and dynamics that need improvement. Therefore, it is essential to acknowledge the limitations inherent in the intervention: assuming loopholes in a method helps to prevent them, making the implementation more complete. The articles from the healthcare sector exhibited, on average, a greater degree of awareness of the limitations of the AI method. Several healthcare researchers discussed a concern that when the emphasis is solely on highlighting the positive changes, the various difficult and distressing situations for participants may be overlooked (e.g., Hung et al., 2018; Knibbs et al., 2012; Magnussen et al., 2019). Pereira et al. (2015) pointed out that participants could be reluctant to share their negative experiences during the AI process, leaving these experiences unarticulated and unsolved. Furthermore, Chauke et al. (2015) reported that over-reliance on positivity during AI stimulated feelings of entitlement in participants who had negative perceptions of their profession, to begin with. This, in turn, made them form even more negative opinions of their profession than before AI. Dewey et al. (2022) studied the experiences of medical residents and contrasted distress (defined as destructive stress) and eustress (stress that leads to positive change). The authors pointed out that it was difficult to clearly delineate between the two types of stress since the AI approach tends to concentrate solely on positivity.
In contrast to healthcare, the studies from the fields of higher education and management invested less effort in recognizing limitations. Still, in higher education, Frerich and Murphy-Nugen (2019) acknowledged the possible criticism of AI for masking oppressive experiences at the discovery stage, and Miles et al. (2018) discussed the critique that fixated focus on positivity leads to this type of research lacking rigor.
In management, two of the reviewed studies mentioned the limitations of AI. Crick and Crick (2016) pointed out that one of the limitations of this approach is related to difficulties in obtaining participation in the AI process. This observation aligns with the viewpoint expressed by Harmon et al. (2012), further reinforcing the significance of this concern. Also, Wall et al. (2017) pointed out that the long-term success of AI depends on factors such as the perceived availability of resources and the positive organizational climate, generated by the AI process.
Some studies from healthcare described steps to counter the AI method’s possible drawbacks. Diedricks et al. (2018) sought to identify barriers the students living with HIV faced in the discovery phase and then looked for desirable goals in the dream phase. Even though Diedricks et al. (2018) did not start with finding things for appreciation, the reported process still helped the students living with HIV discover their strengths and see new possibilities. Watkins et al. (2019) collected descriptions of negative experiences and reframed them as values recognized by the participants. For example, if the participants raised concerns about overcrowding in the emergency department, the researchers re-framed this concern as participants valuing the spaciousness of the room. These examples provide valuable insights into possible ways of addressing the inherent limitations of AI.
Efforts to Further Develop the AI Model
We could not identify any studies that were specifically designed with a goal to further develop and improve the AI model. One may argue that utilizing modified or alternative models can count as an attempt at model development. Indeed, Guo et al. (2022) suggested a model with two added steps, inception and keep, which are likely to enhance the effectiveness of the AI. However, such attempts simply employed a modified approach rather than explaining the advantages that differentiate the new approach from the traditional model. As a result, it cannot be argued that studies utilizing alternative models have demonstrated a significant interest or effort in model development.
Still, there were some cases that can be used as a starting point of future model development efforts. For example, Holtrop et al. (2019) used a rigorous procedure for selecting participants who successfully managed chronic pain and asked them questions aimed at identifying best practices. Additionally, during the discovery phase of projects, reported in Diedricks et al. (2018) and Watkins et al. (2019), participants were asked to share negative experiences as well as positive ones.
Comparison of Study Areas on Themes.
aNumbers that are higher than for the other two areas by 10% or more are shown in bold.
Discussion
This study aimed to review, critique, and synthesize the existing literature on AI in the fields of healthcare, higher education, and management by using the integrative literature review method. In this section, we summarize the major findings of the study and interpret them using the related HRD and organization studies literature as a lens.
First, we found that AI-related research is most actively conducted in healthcare: 62% of the reviewed literature came from the healthcare field, with management and higher education trailing significantly.
Second, we identified multiple purposes for adopting the AI method, and it was observed that the extent to which researchers faithfully follow the 4D cycle varies depending on the objectives. AI was often used to achieve desirable organizational change. In that case, the research generally followed all the steps in the 4D cycle. Additionally, researchers used AI to discover the object of appreciation by looking at past positive experiences. This type of research typically utilizes only the first two steps of the cycle. Some studies adopted AI only as a research tool or used elements of AI as part of their research method toolkit.
Third, we found evidence that AI has positive effects on multiple levels of organization. As pointed out by Fry et al. (2002) the early research on AI was mostly focused on organizational transformation efforts. Our study suggests that this is changing: there is a growing number of research articles, reporting AI’s impact on individuals and at the group level. Overall, our study showed that there is a fair number of studies reporting AI’s positive influence on all three levels: individual, group, and organization.
Fourth, our results suggest that AI can still have positive effects even without following all steps suggested by the 4D cycle. This depends, however, on the goals of the intervention. Thus, if the goal is to start an initial conversation about the needed changes by creating a positive frame of discourse, using just the first two steps could be sufficient. We need to stress, though, that we are making this statement with caution since the number of studies that reported such uses of AI is small.
Fifth, we discovered several innovative applications of AI. In higher education, researchers made multiple attempts to apply AI to program and instructor reviews. This supports an earlier suggestion, made by Preskill and Catsambas (2006) that AI could be incorporated into the evaluation process. We also found that AI has been applied as a teaching method, diagnosis tool, and discussion guide.
In addition, we found multiple instances of using modified AI models. For example, the 4I cycle was such a modified model. However, the differences between the alternative models and the traditional 4D model were not that significant. In most cases, the authors of these alternative models were using different sets of labels, perhaps in an attempt to make the language more descriptive (Preskill & Catsambas, 2006).
A promising direction in developing alternative models of AI could be offered by applications of Critical HRD. Grant and Humphries (2006) suggested that critical theory offers a valuable framework for evaluation, emphasizing its emancipatory aspirations beyond merely exposing exploitation. By highlighting the emancipatory aims of critical theorists, a deeper understanding of complexities, including power dynamics, in the AI process can be achieved to serve participants’ emancipatory goals.
An important finding of our review is that even negative experiences can be successfully incorporated in the AI process. While some AI scholars maintain that the process should focus on the unconditionally positive (e.g., Kadi-Hanif et al., 2014), we found multiple examples of the use of negative experiences to identify positive aspects of any given situation (Diedricks et al., 2018; Holtrop et al., 2019; Watkins et al., 2019).
Our findings also point towards a strong relationship between AI and action research and show how both can interact to amplify each other’s strengths. Some of the reviewed articles used the action research method in their studies, and some considered AI as a form of action research (e.g., Ogude et al., 2019; Watkins et al., 2019). Combining action research, which centers on problem-solving, with AI brings a positive perspective lacking in traditional approaches (Magnussen et al., 2019). In some cases, because of the action research’s emphasis on the “action,” the action plan created by the AI process was more readily implemented without losing momentum, while AI supported the process with an immense supply of positive energy (Falk, 2014). In addition, critical self-reflection facilitated by the action research is expected to enhance reflexivity throughout the AI process, which could also help address limitations of overlooking negative experiences.
Based on our findings, we propose a systems model for understanding the AI process (Figure 2). The purpose of adopting AI is an input, as it acts as a stimulus that triggers the AI application. The application of AI is a process of transformation. According to our findings, AI is generally used following the 4D cycle, but the application of this cycle and steps is expanding and changing. The output of the proposed model includes positive effects (on individual, group, and organization levels), and limitations, which include overlooking or not fully addressing the negative dynamics or processes at all three levels. Systems model of AI process.
Implications for Research
We invite researchers to further study AI to fill the gaps in current research and make improvements in current applications. One such area for future research would be the further development of the AI model. The current model of AI could be developed focusing on two major areas. First, a refined model could clarify the effect of AI on different levels of organization and the interplay of these effects. Multi-level research designs would be applicable to such research. Second, to further refine the current model, it would be beneficial to create a set of criteria that can be used to examine the universe of issues that can be addressed by AI interventions, delineating them from issues that may fall outside the scope of AI interventions.
Propositions for Future Research.
Having access to a list of documented cases of achieving specific, clearly defined outcomes at different levels will be beneficial to HRD practitioners, who could use this information to decide whether to use AI in specific situations. Access to such evidence-based lists or outcomes would also provide a tool for justifying the use of AI to organizational stakeholders. Several studies in our sample discussed various approaches to such assessment of outcomes. Experimental studies were conducted comparing groups exposed to AI with control groups (Mandal, 2022; Scala & Costa, 2014). Jewell and Murphy (2020) compared the before and after data of applying AI to a training program and showcased the numerical growth of the desired outcomes. Therefore, the change in outcome variables was detected based on data collected from the same group, avoiding the difficulties of using the control groups with identical conditions. Jewell and Murphy’s (2020) approach can be applied in a broader range of situations if the researchers can document the state that existed prior to AI application.
When it comes to methodology, we hope that the research methodology toolkit of AI researchers will become more varied. Most reviewed studies utilized qualitative methods, with the bulk of these studies utilizing general qualitative, case study, and action research approaches. Conducting more experimental, quantitative studies, mixed method, and multi-level studies may enrich the findings in AI research. Along the way, the researchers should be attentive to providing clear explanations of their methodological choices.
Implications for Practice
Our findings suggest that HRD practitioners in business and industry could greatly capitalize on the learning from the experience of applying AI in other settings (e.g., in hospitals and in educational institutions). We found numerous examples of such applications outside the traditional business settings, and some of these applications can be considered innovative. Practitioners may consider applying AI for reviewing programs/instructors, conducting diagnoses for learning and development, creating hospitable learning spaces, leading focus group discussions, and so forth, as shown in our findings.
Furthermore, our results serve as a reminder to practitioners that AI’s purpose is not limited to facilitating organizational change. Examples reviewed by us demonstrate that it is successfully used in facilitating transformation on individual and/or group levels by helping individuals or groups to change their mental frames, by discovering new objects of appreciation, and perceiving phenomena in new ways.
Finally, we suggest that HRD practitioners need to reflect on and acknowledge the inherent limitations of AI practice and develop their own approaches to addressing criticism of AI and justifying the use of AI in specific situations. For example, instead of ignoring negative dynamics at the discovery phase, a better strategy would be to document such experiences with the goal of discovering the underlying positive lessons behind those experiences.
Limitations
As with all research, our study had limitations. Our literature search found thousands of English-language peer-reviewed articles that include “appreciative inquiry” in any field. Given the extensive nature of AI-related research, we refined our search criteria to include only articles with “appreciative inquiry” in their titles. This, however, could mean that we missed some relevant articles for this study. We recommend that future research endeavors explore the possibility of broadening search criteria to capture the wider scope of AI literature.
Further, while we expect that the majority of AI-related work is published in English-language publications, it is likely that important related work has appeared in non-English-language publications as well. Therefore, a future literature review-based study of AI could be co-authored by a group of scholars who are conversant in languages other than English. Such research could not only identify the AI literature published in other languages, but also detect potential differences in topics of interest to scholars from various regions of the world.
Footnotes
Declaration of Conflicting Interests
The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Funding
The author(s) received no financial support for the research, authorship, and/or publication of this article.
