The use of Artificial Intelligence for Idea Generation in the Innovation Process
Ode Plätke and Richard C. Geibel
Abstract This presentation contains a systematic literature review, conducted using the PRISMA framework, examines the use of artificial intelligence (AI) for idea generation in the innovation process. To this end, fifteen articles from four databases were identified as appropriate and subjected to descriptive analysis, literature classification, and thematic synthesis as part of the consolidation of results. The focus of the analysis was to identify potential applications and limitations of AI for ideation in order to propose a framework for implementing AI in the innovation process to foster the ideation phase.
The thematic synthesis revealed that current AI models at the time of the research, are best used to support and facilitate human innovators by leveraging their data analytics capabilities to provide human innovators with valuable information or to provide procedural support. Nevertheless, AI models are capable of independently generating new ideas based on combinations or own generations, but this capability is not advanced enough in current models to generate innovative ideas independently and reliably and fully replace human innovators. Therefore, validation by a human user is still recommended, which at the time of this research precludes a full replacement of the human component.
Based on these findings, a hybrid intelligence model is proposed in which AI and human innovators collaborate in the idea generation phase. In this process, AI with its technical capabilities supports the human innovator in its human capabilities maintaining the key position human innovators hold in this process. However, due to the rapid development in the field of artificial intelligence, new research approaches are constantly emerging.
Keywords: AI, Innovation, Idea generation, ideation, hybrid intelligence model
Introduction
The emergence of artificial intelligence has revolutionized a wide variety of sectors, and, through Chat GPT, it is also gaining traction with the general public. As this landscape continues to evolve, new applications of AI are constantly opening up (Mhlanga 2023, n.p.). Thus, the innovative character of artificial intelligence may itself be a driver for innovation. This also emerges from a study by the auditing and consulting firm PwC, which recognizes a value generation through AI for the purpose of innovation of products and services (PwC 2022, n.p.). The starting point of the innovation process is the generation of ideas. It is in this fertile ground that the seeds of potential are sown, germinating into concepts that eventually lead to the aspired innovations (Preez & Louw 2008, p. 9).
Thus, in the wake of this rapid development of AI, it is important to ask what disruptive elements are impacting the innovation process and in what ways Artificial Intelligences are changing how ideas are generated and innovations emerge. From this claim to investigate the role of artificial intelligence for idea generation within the innovation process, the research question underlying this research is derived: How can artificial intelligence be used for idea generation in the innovation process?
This research question suggests the investigation and presentation of the relevance of the subsystem of idea generation in the innovation process, as well as the identification from the literature of possible applications and limitations of artificial intelligence for idea generation, in order to provide a multifaceted view of the topic and to round off the answer to the research question with a conceptual framework for the potential implementation of AI in the innovating process to support the generation of ideas.
Exploring this topic, a Systematic literature review is conducted by examining the research landscape for the object of inquiry and consolidating the thematically congruent contributions to illuminate a more nuanced understanding of the potential use of AI in the innovation process.
For the purpose of creating a compelling roadmap, this introduction is followed by the methodological approach to the Systematic Review with a stringent presentation of the individual steps of the review. After the presentation of the methodological background of the execution, the results of the review are treated on the basis of different synthesis methods. Finally, the results from the synthesis are evaluated in a conclusion.
Theory
The systematic literature review is characterized by a high degree of rigor, emphasizing the importance of validity, reliability, and objectivity. Internal validity is ensured through quality assessments, and reliability is upheld by transparent design, conduct, and analysis (Templier & Pare 2015, p. 113).
The review adheres to the ‘Preferred Reporting Items for Systematic Reviews and Meta-Analyses’ (PRISMA) framework, a widely accepted guide for systematic literature reviews, prescribing the performance of a literature review and the content to be included (Page et al. 2021, p.1).
Methodology
This chapter outlines the methodology used to address the central research question of this thesis. It employs a systematic literature review as a comprehensive approach to develop a conceptual framework for utilizing artificial intelligence (AI) in idea generation within the innovation process.
The research question centers on AI’s role in idea generation within the evolving digital landscape. The review’s objective is to identify and synthesize potential AI applications and limitations in idea generation, ultimately creating a theoretical framework for integrating AI into the innovation process.
Eligibility Criteria
The inclusion criteria specify that articles must be in English, published between 2013 and 2023, peer-reviewed, and thematically relevant to the research question.
Search Strategy
The search strategy relies on a keyword search in four electronic databases: EBSCO, ScienceDirect, Springer Link, and Wiley. The search string combines keywords related to AI, idea generation, and the innovation process using Boolean operators. Resulting in the following search string: (“artificial intelligence” OR “AI”) AND (“idea generation” OR “ideation” OR “design thinking”) AND (“innovation process” OR “innovating process” OR “innovation management”).
Study Selection
A structured approach using the “Rayyan” platform is employed for screening and eliminating duplicates. Records are screened based on title, abstract, and full text to determine inclusion or exclusion, reducing the risk of bias.
Data Extraction
Data is extracted from eligible articles, encompassing both descriptive elements and conceptual evidence to offer a comprehensive overview of the research landscape.
Quality Assessment
A custom consolidated risk of bias assessment matrix is utilized, employing tools such as the Mixed Methods Appraisal Tool for empirical research (Hong et al. 2018), the AMSTAR 2 for review papers (Shea et al. 2017), and the criteria for theoretical conceptual papers by Margherita, Elia and Petti (2022).
Methods of Synthesis
For the purpose of synthesizing the results, after a detailed presentation of the quantitative data of the yielded papers, a descriptive analysis is made of the thematic synthesis of the advantages and limitations of AI for idea generation.
Results
In this section, the results of the conducted literature search performed on the 10th of April 2023, the screening, and the data extraction as part of the review are presented, which was designed according to previously presented methodology.
The search, which included the four databases EBSCO, ScienceDirect, Springer Link and Wiley, yielded a total of 17,093 records, which represented a vast range of potentially relevant sources that could contribute to the review. However, to ensure that the selected literature was meeting the inclusion criteria, the following set of predetermined filters was set in the search masks of the databases to narrow down the scope of the literature: date range between 2013 and 2023, language English, as well as only peer reviewed journal articles.
For the purpose of narrowing down, the databases were filtered according to (“business”) economic reference, if possible. This initial search and filtering procedure, resulted in the exclusion of 15,112 records, leaving 1,981 records eligible for the Screening phase.
The screening process entailed a careful examination of the remaining records’ titles and abstracts within Rayyan, to determine their relevance to the research question. The automated identification and manual reviewing of those potential duplicates resulted in 186 duplicate records, being excluded. Following this assessment, 1,839 records were identified, based on the title and/or the abstract that did not address this thesis research topic or were otherwise unrelated, and thus, were excluded from the review. Being left with 142 reports that warranted a more detailed evaluation to ensure their suitability for inclusion in the synthesis. All 142 reports were successfully retrieved for a comprehensive evaluation, encountering no issues in accessing the full texts of these sources. During this stage, each report was assessed against the predetermined inclusion and exclusion criteria to confirm that they were both relevant to the research question and of appropriate quality. This resulted in 127 reports that did not meet the stringent criteria for inclusion.
The primary reasons for the exclusion of these 127 reports were twofold. First, 111 reports were found to be non- conforming to the subject matter, meaning that they did not directly address the research topic or were peripheral related at best, offering no contribution for the synthesis. Second, 16 articles were excluded on the rationale that the type of the article did not meet the requirements.
Ultimately, after this exhaustive and meticulous screening and selection process, a total of 15 papers was identified as eligible for inclusion in the systematic literature review. These carefully chosen sources provide valuable insights and contributions to a comprehensive understanding of the thesis topic.
The following Table 1 provides a comprehensive overview of all studies that met the eligibility criteria and were therefore included in this review. The Table shows for each of the articles, the year of publication, the authors, the title of the publication, the hosting journal, issue, identifier of the article or the corresponding page numbers within the hosting journal, and the specific contribution each article made to this thesis and the underlying research question.
4.1 Risk of Bias Assessment
The Quality of each Article was assessed using a three-level rating system, represented by the following symbols:
(✓): The criterion is met, indicating low risk of bias or high methodological quality.
(?): It is unclear whether the criterion is met due to insufficient information, indicating unclear risk of bias or uncertain methodological quality.
(X): The criterion is not met, indicating a high risk of bias or low methodological quality.
The risk of bias assessment results predominantly indicates a low risk of bias within the examined articles, reflecting a robust body of research and allowing for a non-discriminatory interpretation of the results. This assessment suggests that most authors have employed sound methodologies and analytical techniques, contributing to the reliability of their respective research findings.
4.2 Thematic Synthesis of Applications
For the thematic synthesis of applications of AI for idea generation the scheme proposed by Thomas and Harden (2008, p. 1) was followed. Therefore, individual value concepts taken from the yielded papers were inductively combined to form higher-level themes. The individual codes were consolidated based on thematic correspondence to other codes. The codes extracted from the articles were checked for commonalities with other codes or for multiple mentions, aiming to infer, inductively, a category superordinate to the concepts that satisfies the classification of common codes. This process resulted in the identification of nine common value-describing concepts related to AI’s role in idea generation, allowing for consolidation into a three-dimensional application space. In this application space, the concepts related to AI’s use for idea generation were integrated into the categories “Informing”, “Promoting” and “Generating” Furthermore, these three dimensions can be abstracted into a bilateral concept comprising the two dimensions of “Human Enablement” and “Autonomous Generator”. For a comprehensive overview of the value propositions and dimensions, please refer to Table 2.
The first dimension, “Inform,” encompasses four value propositions, namely data mining (I1), structuring (I2), data analysis (I3), and pattern recognition (I4). These value propositions underline the diverse ways AI technologies contribute to idea generation. For instance, AI technologies utilizing Natural Language Processing (NLP), such as chatbots, are shown to facilitate information collection and data retrieval from enterprise databases. Similarly, language models like BERT, developed by Google, have proven effective in mining data from social media platforms and identifying solution-oriented content.
The second subcategory within the “Inform” dimension, “Structuring (I2)”, involves the use of neural networks to categorize content and identify unique ideas. This capability is recognized for its ability to enhance idea generation by structuring solution-related content and efficiently identifying knowledge streams. Additionally, semantic categorization of idea content by AI is noted to promote the generation of novel ideas.
“Data analytics (I3)”, the third element of the “Inform” dimension, involves AI’s capacity to store, analyze, and encode large amounts of data. This capability facilitates the exploration of solution spaces, fostering divergent thinking in users to support idea generation. It is highlighted as a valuable aspect for innovation management.
AI-based “pattern recognition (I4)” plays a crucial role in identifying regularities and patterns in user behavior, which can lead to content recommendations tailored to user tastes. This benefit is recognized as a valuable utility aspect for generating ideas and supporting innovation.
Moving to the “Promoting” dimension, a notable category is “Bias Mitigation”. AI’s ability to mitigate cognitive biases and cognitive fixation on prior ideas is identified as a significant benefit. It is viewed as a means to reduce barriers in idea generation and stimulate user creativity. This aspect is appreciated as a valuable opportunity for AI in promoting creative thinking and problem-solving.
AI’s role in stimulating creativity is another key element in the “Promoting” dimension. Various methods, such as presenting visual stimuli and generating stimulus ideas, are discussed as ways AI can inspire and enhance creative thinking. Conversational agents and language models like ChatGPT are mentioned as valuable sources of inspiration and ideas.
The third category within the “Promoting” dimension is “Process Support (P3)”, which addresses AI’s role in supporting the creative process by providing targeted guidance, communication facilitation, and comprehensive moderation. AI’s ability to guide users toward novel ideas and enhance the efficiency and effectiveness of ideation efforts is highlighted.
Finally, the “Generating” dimension encompasses two value propositions: “Combination (G1)” and “Generation (G2)”. AI’s capacity to combine existing ideas to form novel ones is discussed, emphasizing the potential of AI in enhancing creative thinking through idea synthesis. Additionally, AI is recognized for its ability to independently generate new ideas, speeding up the idea generation process and providing a diverse range of novel ideas.
“Data analytics (I3)”, the third element of the “Inform” dimension, involves AI’s capacity to store, analyze, and encode large amounts of data. This capability facilitates the exploration of solution spaces, fostering divergent thinking in users to support idea generation. It is highlighted as a valuable aspect for innovation management.
Already some real-world examples showcase AI’s effectiveness in generating ideas, such as automatically generating fragrance formulas, performing various tasks like idea generation, and continuously producing ideas when prompted (Raisch & Krakowski 2021, p. 196).
Thus, regarding those value propositions, the role of AI in idea generation in its multiplicity is thoroughly explored, encompassing various benefits and applications. These findings highlight the extensive potential of AI in enhancing the creative process and supporting users and innovators in generating valuable and innovative ideas. Nevertheless, the role of the human being cannot be ignored and many of the value prepositions emphasize the use in support of the human being.
4.3 Thematic Synthesis of Limitations
Similar to the synthesis of AI applications for idea generation, a thematic synthesis was employed to identify and analyze limitations to AI in the context of ideation. The articles were scrutinized for evidence of constraints and challenges associated with AI’s role in idea generation, which were then distilled into value propositions and subsequently organized into broader dimensions. Among the fifteen articles included in the review, limitations of AI in idea generation were identified in nine publications. These limitations were encapsulated in eleven value propositions within a coding framework, and these value propositions were further categorized into three overarching dimensions. An overview of these value propositions and the higher-level themes is presented in Table three.
4.3.1 Implementation
The first dimension, “Implementation”, comprises value propositions related to the practical challenges of applying AI in innovation management. These include “Application“, “Accessibility” and “Technology Readiness.”
- Under the "Application" value proposition, it is noted that AI-based innovation management does not lend itself to universal application. Instead, tailored approaches are required to address specific innovation challenges, rejecting the one-size-fits-all approach. This highlights the need for adaptive and flexible AI systems that can cater to unique contexts and requirements“
- "Accessibility" is another dimension where challenges are evident. Unstable internet connections are identified as a hindrance to the accessibility of AI, emphasizing the dependency on robust and stable infrastructure. This limitation brings attention to the digital divide, which could restrict the reach and effectiveness of AI in innovation management.
- The "Technology Readiness" value proposition points out that the technological capabilities of AI systems remain relatively limited, with few systems meeting the desired level of readiness. The gap between the theoretical potential of AI and its practical usability in real-world scenarios is underscored. The maturity of natural language processing mechanisms is also deemed insufficient for human-like conversations with chatbots, highlighting the preference for human communication partners.
4.3.2 Capabilities
Moving to the second dimension, “Capabilities”, various value propositions shed light on AI’s limitations in performing specific functions or achieving desired outcomes. These include “Automation“, “Exploration“, “Creativity“, “Autonomy” and “Human Interaction.”
- Concerning "Automation", AI solutions aim to automate workflows, but it is recognized that an entire set of tasks cannot always be fully automated. This implies the need for human involvement and oversight in processes that are too complex or nuanced for AI to handle independently.
- In the domain of "Exploration", AI systems are noted to have limited ability to explore problem spaces, a crucial function often required in innovation management. The inability of AI to explore uncharted territories and acquire new knowledge presents a limitation in innovation contexts.
- The "Creativity" value proposition highlights that AI struggles to emulate human creativity, a central aspect of innovation management. Given the importance of creativity in innovation, this limitation is of particular significance“
- "Autonomy" refers to AI systems' capacity to operate independently. These systems often struggle to navigate solution spaces independently, lacking self-awareness and purposefulness. As a result, human involvement in specifying goals remains essential in innovation management.
- Under the "Human Interaction" value proposition, issues may arise in human-computer interaction. AI, such as ChatGPT, may struggle to understand certain nuances or contexts, and ethical and cultural factors may be challenging to grasp. These limitations point to potential shortcomings in AI's ability to fully mimic human interactions.
4.3.3 Quality
The third and final dimension, “Quality”, includes value propositions like “Reliability“, “Data Security” and “Training Data.”
- The "Reliability" value proposition raises concerns about the trustworthiness of AI systems and the quality of their results. It is noted that the quality and context of AI-generated ideas should be rigorously tested, as AI can introduce statistical biases and limitations. Understanding the results of AI, particularly those generated by pattern-based methods, can be challenging, emphasizing the need for human validation“
- "Data Security" is another critical dimension, where issues related to privacy, security, and potential misuse of data are highlighted. These concerns underscore the importance of addressing data security when using AI systems in innovation management.
- The "Training Data" value proposition underscores the significance of the data used to train AI systems. Insufficient labeled training data and data quality issues can hinder AI's effectiveness. Bias in training data can potentially harm language models. High-quality, representative datasets are deemed essential for optimal AI performance.
Summing up the findings reveal a wide range of potential challenges that need to be considered for the effective use of AI for idea generation. These limitations include practical implementation issues, capability constraints and concerns about the quality and reliability of AI systems. Addressing these limitations is essential if the full potential of AI in innovation management is to be realized.
Conclusion
This research delved into the role of Artificial Intelligence (AI) for generating ideas, a pivotal aspect of the innovating process. By conducting a systematic literature review of 15 selected articles, it becomes evident that AI holds immense potential to support idea generation, offering numerous applications but also acknowledging certain limitations.
The review highlights the capabilities of AI systems not only to generate ideas independently but even more to assist in enabling human capabilities by fostering creativity, processual support, and data processing. However, the research uncovers also substantial challenges linked to AI implementation in idea generation. Issues concerning data quality, biases, and the opacity of AI systems may act as potential barriers to successful integration. Therefore, the aim should be to offer transparency regarding AI’s current landscape, enabling informed decision-making and mitigating potential disadvantages
The reviewed articles underscore the significant potential for AI to generate autonomously and through combination but even more to enhance human innovators in the ideation process by various applications. This synergy between human abilities and artificial intelligence can form an even more valuable instance that can be referred to as a “hybrid intelligence”. Since AI is capable to complement human capabilities rather than replacing them. And still, if AI independently generates ideas, the importance of human validation is emphasized nevertheless, as suggested by Dwivedi et al (2023).
Thus, this research shows that AI, with its multiple capabilities, can support ideation in many ways by forming a symbiotic entity with human users to transform the ideation process in innovation management and unlock new potential. However, consideration of the limitations and the correct and thoughtful use of AI are essential for success.
Given the dynamic nature of AI development, this research serves as a snapshot of the evolving landscape. It points towards numerous opportunities for further research on AI’s role in the innovation process and the interaction between AI and human innovators.
References
Bouschery, S. G., Blazevic, V. & Piller, F. T., 2023, ‘Augmenting human innovation teams with artificial intelligence: Exploring transformerbased language models’, Journal of Product Innovation Management, 40(2), 129-153
Clegg, M., Hofstetter, R., Blohm, I. & Bravin, M., 2022, ‘Human-Machine Creativity: How AI Can Influence Human Creativity in Open Innovation’, Marketing Review St. Gallen, 6, 40-47.
Dkhil, A., Dammak, H., Mouatassim, M. & Gardoni, M., 2022, ‘A new approach to ideation based on linkography combined with graph theory’, International Journal on Interactive Design and Manufacturing, 16(2), 627-642.
Du Preez, N. & Louw, L., 2008, ‘A framework for managing the innovating process’, PICMET ’08 – 2008 Portland International Conference on Management of Engineering & Technology, Cape Town, South Africa, 546-558.
Dwivedi, Y. K., Kshetri, N., Hughes, L. Slade, E. L., Jeyaraj, A. & Kar, A. K. et al., 2023, “So what if ChatGPT wrote it?” Multidisciplinary perspectives on opportunities, challenges and implications of generative conversational AI for research, practice and policy’, International Journal of Information Management, 71, 102642.
Füller, J. Hutter, K., Wahl, J., Bilgram, V. & Tekic, Z., 2022, ‘How AI revolutionizes innovation management – Perceptions and implementation preferences of AI-based innovators’, Technological Forecasting and Social Change’, 178, 121598.
Garbuio, M. & Lin, N., 2021, ‘Innovative idea generation in problem finding: Abductive reasoning, cognitive impediments, and the promise of artificial intelligence’, Journal of Product Innovation Management, 38(6), 701- 725.
Gimpel, H., Graf-Seyfried, V., Laubacher, R. & Meindl, O., 2023, ‘Towards Artificial Intelligence Augmenting Facilitation: AI Affordances in Macro-Task Crowdsourcing’, Group Decision and Negotiation, 32(1), 75-124.
Haddaway, N. R., Page, M. J., Pritchard, C. C., & McGuinness, L. A. 2022, ‘ PRISMA2020: An R package and Shiny app for producing PRISMA 2020-compliant flow diagrams, with interactivity for optimised digital transparency and Open Synthesis Campbell Systematic Reviews’, 18, e1230. https://doi.org/10.1002/cl2.1230
Haefner, N., Wincent, J., Parida, V. & Gassmann, O., 2021, ‘Artificial Intelligence and innovation management: A review, framework, and research agenda’, Technological Forecasting and Social Change, 162, 120392.
Hong, Q. N., Pluye, P., Fabregues, S., Bartlett, G., Boardman, F., Cargo, M. et al., 2018, ‘Mixed Methods Appraisal Tool (MMAT)’, McGill University, Montreal.
Johannsen, F., Schaller, D. & Klus, M. F., 2021, ‘propositions of chatbots to support innovation management process’, Information Systems & e-Business Management, 19(1), 205-246.
Margherita, A., Elia, G. & Petti, C., 2022, ‘What is Quality in Research? Building a Framework of Design, Process and Impact Attributes and Evaluation Perspectives’, MDPI, 14(5), 3034
Mariani, M. M., Machado, I. & Nambisan, S. 2023, ‘Types of innovation and artificial intelligence: A systematic quantitative literature review and research agenda’, Journal of Business Research, 155(B), 113364
Mhlanga, D., 2023, ‘The Value of Open AI and Chat GPT for the Current Learning Environments and the Potential Future Uses’, SSRN, 09.05.2023.
MIT-IBM Watson AI Lab, n.y., Research, viewed 11.05.2023, from https://mitibmwatsonailab.mit.edu/research.
Page, M. J., Moher, D., Bossuyt, P. M., Boutron, I., Hoffman, T. C., Mulrow, C. D. et al., 2021, ‘PRISMA 2020 explanation and elaboration: updated guidance and exemplars for reporting systematic review’, The British Medical Journal, 372 (160).
Poser, M. Küstermann, G. C., Tavanapour, N. & Bittner, E. A. C., 2022, ‘Design and Evaluation of a Conversational Agent for Facilitating Idea Generation in Organizational Innovation Processes’, Information Systems Frontiers, 24(3), 771-796.
PwC, 2022, ‘PwC 2022 AI Business Survey’, viewed 10.07.2023, from https://www.pwc.com/us/en/techeffect/ai-analytics/ai business-survey.html .
Raisch, S. & Krakowski, S., 2021, ‘Artificial Intelligence and Management: The Automation-Augmentation Paradox’, Academy of Management Review, 46(1), 192-210
Shea, B. J., Reeves, B. C., Wells, G., Thuku, M., Hamel, C., Moran, J. et al., 2017, ‘AMSTAR 2: a critical appraisal tool for systematic reviews that include randomized or non-randomised studies of healthcare interventions, or both’, The British Medical Journal, 358, j4008.
Templier, M. & Paré, G., 2018, ‘A Framework for Guiding and Evaluating Literature Reviews’, Communications of the Association for Information Systems, 37(6), 113-137.
Thomas, J. & Harden, A., 2008, ‘Methods for the thematic synthesis of qualitative research in the systematic reviews’, BMC Medical Research Methodology, 8, 45.
Verganti, R., Vendraminelli, L. & Iansiti, M., 2020, ‘Innovation and Design in the Age of Artificial Intelligence’, Journal of Product Innovation Management, 37(3), 221-227.
Wahl, J., Hutter, K. & Füller, J., 2022, ‘How AI-Supported Searches Through Other Perspectives Affect Ideation mOutcomes’, International Journal of Innovation Management, 26(9), 2240028.
Zhao, T., Yang, J., Zhang, H. & Siu, K. W. M., 2019, ‘Creative idea generation method based on deep learning technology’, International Journal of Technology and Design Education, 31(2), 421-440.