Assessing the state of biologically inspired design from three perspectives: academic, public, and practitioners

Biologically inspired design (BID) applies natural solutions to engineering challenges. Due to the widespread success of BID, we examine the following research question: how does the purpose of applying, the inspiration source, and the application of BID differ between academics, the public, and practitioners? Answering this question can help us design the tools used to support BID, provide an understanding of the current ‘state of BID’ and identify where BID solutions have not been widely utilized. Identifying gaps in utilization could prompt investigations into BID methods in new fields. To answer this research question, 660 BID samples were gathered equally from three data sources: Google Scholar, Google News, and the Asknature.org ‘Innovations’ database. The data were classified across seven dimensions and 68 subcategories. The conclusions of our research deliver insights into three areas. First, we identify trends in BID independent of source. For example, 72.5% of the biomimicry samples had the purpose of improving functionality and 87.6% of the samples impacted the usage phase of a product’s life cycle. Secondly, by examining the distribution of BID within each source, we identify areas for potential outreach or application. Finally, by contrasting BID results between three sources (academic, news, and practical case studies) we gain an understanding of the disparities between the three. This analysis provides BID researchers and practitioners with a useful insight into the present state of this field, with the goal of motivating future research and application.


Introduction
Biologically inspired design (BID) is a form of design by analogy that applies mechanisms found in nature to improve human technology , Fu et al 2014, Helms et al 2021. For example, a wall-crawling robot uses contacts inspired by gecko feet and new models of cyber-physical systems that incorporate biologically-inspired objective functions outperform traditional approaches (Keller 2018, Li et al 2022. BID is an acknowledged source of new solutions, and the application of BID in recent years has been increasing steadily. Academic papers and patents related to BID increased by over 500% from 2003 to 2015 (Lenau et al 2018), but there are still significant areas that do not use BID (Hamidouche et al 2019).
Despite a history of successful implementation, substantial barriers to BID are still present. BID is often only used to implement easily accessible analogies that the designer is already familiar with. Thus, tools such as BioTRIZ, BioThesaurus, BioScrabble, Intelligent Biologically Inspired Design (IBID) Interactive System, BioM innovation database, and the AskNature.org database have been developed to enable BID (Vincent et al 2006, Shu et al 2011, Wiltgen et al 2011, Fu et al 2014, Jacobs et al 2014, Goel et al 2022. The design of, and support for, these tools can be improved by understanding how BID is currently applied Weissburg 2007, Shu et al 2011). While some researchers have surveyed specific applications for BID (e.g. networking or manufacturing) (Dressler and Akan 2010, Reap and Bras 2014, Oh et al 2017, Hamidouche et al 2019, no review has attempted to survey how biomimicry is employed overall. Motivated by this gap, the key research question examined in this paper is 'How does the purpose of applying, the inspiration source, and the application of BID differ between academics, the public, and practitioners?' These three perspectives use different lenses to view BID. Academic research explores what type of BID solutions are possible. The public perspective provides insight into what innovations are newsworthy or of interest. Practitioners are those who successfully implement BID solutions in real-world products and services. Practitioners must balance not only what is technically possible, but also what is economically feasible. To investigate this question, a dataset of 660 BID samples was collected and classified across seven dimensions divided into 68 subcategories.
The main contributions of this work are to: (1) Provide insight into the overall state of BID when considering the three perspectives. This was accomplished by combining the three sources and examining for overall trends. (2) Recognize which subcategories (e.g. different solution strategies) are under-represented or over-represented within each of the three perspectives. This contribution can be used by practitioners to identify areas where biomimicry could be more powerfully applied, and by BID tool designers to target under-represented domains.
(3) Uncover mismatches between the three perspectives. This contribution provides an indication of the differences between how these groups view or apply BID. General recommendations are made depending on the type of mismatch. For example, one finding is that inspirations based on reptiles are under-represented in the academic literature, indicating that this could be an area where public interest supports additional research (potential funding implication). Future work can focus on testing the recommendations presented in this article.
The remainder of this paper is organized as follows. The background provides a discussion of BID and compares our approach to previous efforts to survey BID implementation. The methods section lists the steps taken to gather, classify, and analyze the BID samples. The results section details the outcomes for each category and source. Finally, the contributions are highlighted along with future work.

Biologically inspired design
BID transfers solutions from nature to engineering challenges (Yen and Weissburg 2007, 2012, Helms et al 2009, Fu et al 2014. There are two standard approaches for applying BID: problem-driven and solution-based . Problem-driven BID seeks a biological solution for a previously acknowledged technical problem (Shu et al 2011, Nagel et al 2014. Solution-based BID begins with the discovery of a biological phenomenon that is applicable to an engineering issue. An example of a solution-based BID is Velcro®. George de Mestral was inspired by the discovery of burdock seeds that clung to his dog's fur (Shu et al 2011). Although some studies have reported that problem-driven BID is more commonly utilized than solution-based BID (Yen and Weissburg 2007, Shu et al 2011, other studies have observed that they are used at an approximately equal rate among novices (Weissburg et al 2010). More information on BID theory can be found in (Helms et al 2009, Lenau et al 2018. BID demands a comprehensive knowledge of both biological and engineering principles (Yen and Weissburg 2012). The challenge of combining these fields is that biologists and engineers follow separate methodologies, and therefore convey information differently . For instance, biologists are often not trained in design approaches (Yen and Weissburg 2012). These differences make it difficult to combine ideas from biology and engineering (Helms et al 2009, Yen andWeissburg 2012). Locating appropriate biological analogies is also difficult (Shu et al 2011), and the ability to recognize a suitable source of biological inspiration demands a high level of expertise in both design theory and biology (Yen and Weissburg 2007).

Previous BID surveys
Many previous BID surveys focused on a specific application domain. For example, Oh et al examined BID applications within multi-robot pattern formation (Oh et al 2017). Similarly, Dressler and Akan focused on the usage of BID in networks (Dressler and Akan 2010). In another review of BID, Reap and Bras provided methods for extracting principles from biology and ecology literature to create bio-inspired guidelines for environmentally benign design and manufacturing. The authors were primarily focused on sustainable engineering and did not analyze BID outside the scope of this area of engineering (Reap and Bras 2014). Chirazi et al's survey primarily focuses on how to effectively apply BID in industry. These authors provided insight into how to best utilize BID, but did not detail where it is and is not commonly used (Chirazi et al 2019). Providing a broader view of BID, Fu et al, analyzed BID as a form of design-by-analogy and identified where BID can be further explored in future research (Fu et al 2014). Their analysis, however, did not survey the application of BID. Each of the previous surveys provides insight into BID, but unlike our focus, they do not provide a broad review of current BID employment while considering different perspectives.
Two BID reviews, however, go through a similar process as this paper. Jacobs et al looked at 380 examples of biomimicry to gain an understanding of the global state of BID (Jacobs et al 2014). The authors provide a perspective on BID on a broad scale. Their four points of focus were to determine if products labeled as BID were true examples of BID, if there were patterns in the BID design practices that can be learned from practitioners, how much BID products exploit the abilities of biological systems, and to what extent BID examples mimic nature. As in this paper, Jacobs et al built a database of BID examples and evaluated each example with a set of parameters and questions. The authors examine the quality of the BID examples and the design processes, whereas we focus on BID application, inspiration source, and purpose. Jacobs et al also do not examine the role of perspective. A similar review was performed by Sartori et al. That review also examined a set of samples and the knowledge transfer that occurred, but their work only examined 20 BID examples through the SAPPhIRE model of causality. The goal of Santori et al's work was to provide a general design process and guidelines to help with idea transference in biomimetics, not provide insight into the current state of BID (Sartori et al 2010).

Methodology
The methodology section describes how the BID samples were collected, the classifying procedure, and the definitions used for classification.

Research procedure: data collection, classifying, and analysis
This research consisted of eight steps (figure 1).
Step 1: Identify data sources: the first research task was to identify three sources that would each correspond to a perspective (academia, public, and practitioners). Google Scholar, a database that enables scholarly literature search and retrieval, and provides insight into BID in academic research. The researchers also considered other databases such as the campus library and Web of Science. Google Scholar was selected due to producing more results per search term and to reduce barriers to repeatability.
The second source (for the public perspective) was Google News, which is a database of online news articles from multiple sources. As of January 2023, Google News was the fifth most visited news website in the United States. It was the most visited news website that compiled articles from multiple publications (Majid 2023). As an additional benefit, because the first two sources are both created by Google, we minimize the probability that the underlying sorting algorithm will impact our overall results when comparing sources.
For practitioners, AskNature.org was selected. Practitioners are those that successfully utilize biologically inspired solutions in the real world. Because AskNature.org was created by the Biomimicry Institute to provide examples of successful BID, it was selected as the third source (Shu et al 2011). AskNature.org is a prominent BID resource, with over 200 000 page views per month in 2021 (Ratter 2021). For this study, we utilized AskNature.org's 'Innovations Database,' a collection of case studies designed to highlight BID in use (AskNature 2022b).
Step 2: Define classification approach (dimensions and subcategories): We classify BID across dimensions and subcategories (table 1). Dimensions are over-arching categories that capture how BID is utilized or inspired. The subcategories classify the BID samples within each dimension. For example, within the dimension of application, a sample could be classified in the subcategory software. Full subcategory definitions are given in section 3.2 but are presented here to provide context for the remainder of section 3.1. Our classifying process was inspired by how Reap and Bras classified their sources and information in their review of BID guidelines for manufacturing (Reap and Bras 2014). Reap and Bras sorted their collected examples into separate categories (e.g. biodiversity and competition) to allow for further analysis. We, however, used more detailed dimensions and subcategories than Reap and Bras while also considering the three perspectives.
Dimension and subcategory identification was done using two passes. First, dimension and subcategory identification occurred through a brainstorming session between the authors where subcategories of potential interest were identified based on author experience from previous research (Jastrzembski et al 2021). The brainstormed subcategories were checked against external sources to ensure no subcategories were overlooked. For example, the United Nations sustainable development goals were reviewed after generating the subcategories for Purpose. Next, these initial subcategories were refined by collecting and classifying a small data subset of 60 BID samples. The subcategories were then refined as the authors identified gaps or ambiguities that arose when classifying using the initial subcategories.
The two-pass approach resulted in seven dimensions and 68 subcategories (table 1). These dimensions provide a broad scope of both the BID problem space (what aspects of technology are being addressed) and the solution space (the systems and organisms that provide potential ideas). Both problem and solution spaces are used to guide the   . Together these seven dimensions cover the reason for the application (purpose), how the inspiration was used (application, strategy, and life cycle assessment (LCA) stage), and what provided inspiration for the design (inspiration taxonomy, biome, and mobility). Detailed selection guidance and descriptions of these dimensions and subcategories are presented in section 3.2. The dimensions and subcategories were created by the authors, and they were not duplicated from another source. Different literature references and databases were utilized to brainstorm subcategories and dimensions (see section 3.2), but none were directly copied or used as a primary source of inspiration for dividing up the dimensions and subcategories with the exception of LCA Dimension.
Step 3: Identify search terms: Next, a set of search terms were selected for a Google Scholar and Google News sample collection. AskNature did not require search terms because we used their entire 'Innovations' database. The term definition began with Shu et al's review, which identified 20 different terms used in conjunction with BID (Shu et al 2011). These terms were combined with 20 additional terms that the authors identified as referring to BID. Together, these 40 terms comprise all the roughly synonymous terms commonly used for BID, as well as domain-specific terms that incorporate biology (e.g. biophysics and biomedical).
Step 4: Define representative sampling approach: A total of 660 samples were collected from the three sources. The AskNature.com 'Innovations' database contained 223 examples of BID at the time of data collection (2 July 2022); therefore, 220 samples were selected for each source. There were more than 220 available samples of BID in both Google News and Google Scholar; therefore, it was necessary to select a representative subset of samples from these two sources. We used a weighted sampling approach for each term in table 2. This approach was necessary to ensure the number of samples classified per term was proportional to its prevalence in either the news or academic literature. A slightly different approach was required for Google News and Google Scholar.  . (1) The total number of Google Scholar search results (N all terms ) generated by the search terms in table 2 is approximately 22 million articles. This analysis was performed on 18 November 2022. The result of this weighting is that 15 of the 40 terms for Google Scholar require no samples (i.e. X = 0). Appendix A provides the number of results required for each search term per equation (1).
A different approach was required to sample Google News, because Google News only reports the first 100 relevant queries. Thus, the results of the Google News search cannot be used to ensure a representative sample for each search term. Instead, Google Trends was used for insight. Google Trends provides an interest score from 0 to 100 for each search term. Google accounts for approximately 60% of US search traffic, providing a reasonable surrogate for public interest in various search terms (Trielli and Diakopoulos 2019). The interest score from 2004 to 2021 was downloaded (reported by Google monthly) and the average was taken. These averages were then used in equation (1) to perform the weighted data collection for Google News. The result of the weighting for Google News is that 26 of the 40 terms for Google Scholar require no samples (i.e. X = 0). Appendix A provides the number of results required for each search term. Google Trends combines interest from Google Search, Google News, and YouTube (Google 2022b), thus the impact of the other two sources could influence our results (for example, if a term was queried more in Google than appeared in the news). We consider this acceptable, however, because Google Search and YouTube also reflect the public perspective.
Step 5: Collect 660 samples for analysis: Samples were collected as per steps 3 and 4. To ensure that previous user activity did not influence Google Scholar or Google News results, the researcher selected had not previously used Google and used a new browser without user history. This factor is most important for Google News, which uses data from a user's online activity along with the key terms used in the search bar to provide relevant news articles (Sullivan 2009). Google Scholar does not use past searches to identify relevant documents (Google 2022a).
The process of screening search results for BID samples was informed by Silva and Braga (2020), who recommended defining a series of criteria that must be met for a paper to be included. We first evaluated whether the sample met three criteria: shows the use of design by analogy (transferring biological inspiration), identifies a specific biological solution, and applies the results to an engineering (human artifact) problem. The use of multiple search terms resulted in duplicates within a single source (e.g. the same article appearing twice in Google News). Each article was only added to the database once, although separate news articles concerning the same BID application were included. Additional news coverage would impact the public's perception of how BID is used, which relates to our research question on how BID is employed. Additionally, certain BID samples appeared in multiple search sources (i.e. appearing in AskNature, Google Scholar, and Google News). We expect some overlap because a lack of overlap would indicate no shared BID examples between academics, the public, and practitioners. Finally, articles from Google News sometimes contain multiple samples of BID. These were recorded in the dataset as separate BID examples. The purpose of this examination is to look at 220 BID samples presented to the public, not 220 articles. News articles are often themed, however, so this could reduce the variety of BID samples in Google News.
Step 6: Classify collected samples: After gathering 660 samples, the first author classified each BID sample across each subcategory using the definitions in section 3.2. Classification is the process of identifying which of the subcategory(ies) apply to each sample when considering each dimension. An example of a classified BID entry is given in table 3. The abstract of the academic article, the information in the AskNature Innovation database, and the full news article were used to make classifying determinations. To ensure that author bias did not influence the classifying results and that the subcategory definitions were sufficiently detailed, inter-rater reliability checks were performed on the 60-sample subset used for the first-pass subcategory identification in step 2. The inter-rater agreement rate was 97%. Disagreements between the raters were discussed, resolved, and the subcategory definitions provided in section 3.2 were updated to reflect the new consensus.
This classifying process was inspired by previous reviews of BID and prototyping (M. E. Reap and Bras 2014, Helms et al 2021, Murphy et al 2021. Helms et al utilized a similar basic classifying scheme to evaluate designs in their paper. For example, some of their overarching categories were titled 'Biome,' 'Structure,' 'Organism,' and 'Function' , which represent common ways biological systems are described, and therefore ways this information can be accessed. In this paper, we are also looking at the biomes and types of organisms that inspired BID. Some samples were classified into multiple subcategories (see section 3.2). For example, a generative design system inspired by both slime molds and human bones (Autodesk and AskNature 2016) would fall into the Inspiration-taxonomy subcategories of both Mammal and Slime Mold. The fully classified dataset is available from the authors by request.
Step 7: Perform statistical analysis: After classifying, step 7 focused on identifying how BID is currently applied and detecting any differences between sources.
Contribution 1 (overall): Occurrence frequency was calculated to provide an indication of overall prevalence.
Contribution 2 (differences within sources): To determine if the observed distribution within a source could be the result of an underlying uniform distribution, the Chi-squared test (χ 2 ) with a type I error rate of 5% (p < .05) is performed. Recall where O is the observed percentage in each subcategory and E is the expected percentage in each subcategory. The χ 2 test also provides insight into which subcategories are most skewed away from the uniform distribution by examining (O−E) 2 . Of note, the χ 2 test is only performed for the dimensions that do not allow a single sample to be classified into multiple subcategories (see section 3.2). For dimensions that do allow for a single sample to be classified into multiple subcategories, the standard deviation of all the calculated subcategory percentages is used as a measure the distribution of the data. Subcategories whose number of recorded samples per subcategory were greater than a standard deviation from the mean of the subcategories for that perspective was defined as elevated or depressed. Contribution 3 (differences between sources): First, the mean number of samples (SAM) and standard deviation (σ) for each subcategory between the three perspectives was calculated. Then, each Table 4. Summary of dimension details >1 subcategory column indicates if a single sample could be classified into more than one subcategory (e.g. crawling and walking for inspiration-mobility), number of subcategories reports the total number of subcategories within each dimension . source's subcategory percentage was compared to the mean of the same subcategory across the three sources. Sources whose percentage per subcategory was greater than a standard deviation from the mean of the three sources was defined as elevated or depressed. Likewise, the χ 2 test was also repeated to analyze the distribution for the same subcategory between sources (assuming uniform distribution).
The average value of the number of samples per subcategories across the three sources was used for the expected value (E) for the χ 2 test.
Step 8: Interpret statistical findings: The results of step 7 were reviewed in the light of current literature on BID practice and design-by-analogy to see if there were any existing theories that might help explain the statistical findings.

Definitions used for bid classifying
When classifying the collected BID samples, the following definitions were used for the seven dimensions and 68 subcategories. These dimensions and subcategories were determined by the authors through a two-pass approach that included classifying a small dataset of 60 BID samples (section 3.1 Step 2). A detailed description of each dimension and subcategory follows, but a summary of dimension details is shown in table 4.

Applications
This dimension was divided into seven subcategories of where researchers have applied BID. Samples could only be classified into one subcategory.
1. Static physical object-: A physical application of BID that utilizes the stationary physical attributes (i.e. does not incorporate moving elements) and applies them to an object. One stationary physical attribute is morphology. For example, the physical structure of the kingfisher's beak was used to make the 500 Series Shinkansen trains more aerodynamic (Frangoul 2020a). 2. Dynamic physical object: A physical application of BID that utilizes the dynamic physical attributes of nature (i.e. how movement occurs in nature) and directly applies them to a moving object. For instance, researchers have taken inspiration from a dog's legs to create better robotic jumping mechanisms (Hyon and Mita 2002). 3. Software: BID is applied to design computer software or algorithms. An example of a software BID application is a computer's response to malware that mimics the human immune system (Kephart 1994). 4. Chemical: An application of BID to make or modify a chemical. This subcategory includes many of the molecular BID applications. For example, synthetic enzymes are inspired by DNA (Schlosser and Li 2009). 5. Network Design: An application of BID to improve a network's design. Networks consist of elements that exchange energy, material, money, or information. This subcategory is different from the Software subcategory because it specifically examines how individual elements are connected and the flows between them. Network design is not limited to information flow. An example of this type of application is insight from the human immune system and central nervous system to create an efficient and unique network (Hoffmann 1986). 6. Process: A natural sequence of events is mimicked.
Process is different from Software because it is not confined to executable programs. An example is a BID toilet, which instead of requiring plumbing, mimics how plants pull water up from the soil (AskNature 2020b). Process is distinct from Static or Dynamic Physical Objects because the object's shape, form, or attributes that create motion are not biologically inspired.
The next three dimensions provide insight into where innovators look for inspiration in nature.

Inspiration-taxonomy
In this dimension, the taxonomy of the source of inspiration is identified as one of 27 subcategories (table 1). For all sources of inspiration in Kingdom Animalia, the creatures were classified into their phyla. Vertebrates, which fall into the phylum Chordata were classified into their subphyla: Mammal, Amphibian, Reptile, Bird, or Fish. For example, robot designs inspired by the characteristics of the modular caudal fin of fish were classified into the subcategory Fish (Kopman and Porfiri 2013). Phyla was not used for inspiration outside of Kingdom Animalia because that would add an additional 75 subcategories, which would reduce the statistical power of our analysis (e.g. a single subcategory would have at most one or two samples) and would not provide useful information for most designers. Thus, samples outside of Kingdom Animalia were classified as nematodes, cnidarians, mollusks, arthropods, annelids, platyhelminths, echinoderms, poriferous, plants, molecules, fungi, enzymes, amoeba, cells of multicellular organisms, bacteria, tardigrades, slime molds, or plankton. Although plankton could be broken down further (e.g. diatoms, crustacean, bacteria), most sources did not provide any additional specification about the type of plankton used as a source of inspiration. Additionally, a subcategory for single-celled organisms such as protozoa could have been added, but there were no examples found in the data. The National Center for Biotechnology Information's Taxonomy Browser tool was used (www.ncbi.nlm.nih.gov/taxonomy) to classify each sample. Finally, some sources of inspiration were not organisms so the subcategories of enzymes, virus, ecosystems, and natural phenomena (e.g. the shape of waves) were used in these cases. Increasing the level of detail in the subcategories could be examined in future research targeting biologists.

Inspiration-biomes
In this dimension, the biome from which the source of inspiration originated was identified. The 'Worldwide/unknown' subcategory is for sources of inspiration found in more than three biomes, if the biome is not able to be determined, or if the source of inspiration is an extinct creature. Samples could be classified into more than one subcategory.

Inspiration-mobility
In this dimension, the mode of movement for the source of inspiration was identified as static, walking, flying, crawling (e.g. snakes), climbing (e.g. traveling up or down a ⩾90 • slope), swimming, microscopic Movement, or N/A. The microscopic movement subcategory is for sources of inspiration that are too small for the eye to see and that does not fall into one of the other subcategories. Examples of microscopic movement include ciliary or flagellatory movement. The N/A subcategory is for sources of inspiration such as natural phenomena or ecosystems. Samples could be classified into more than one subcategory (e.g. swimming birds).

Strategy used to address design issue
This dimension describes the strategy used to solve a design problem, which reflects how BID is applied to improve a device or process. Similar subcategories have been analyzed in other studies. For instance, Vincent et al defined categories such as structure, energy, and substance to understand how biomimetics is applied (Vincent et al 2006). Of note, the Strategy dimension differs from the Application dimension. Strategy captures how a problem was solved, while Application records the type of solution that was created. For example, an Application-Chemical (new chemical) can be made using a Strategy-Procedure (steps to create the chemical). Samples could only be classified into one subcategory.
1. Energy: Designs using energy to solve a problem.
Examples include harvesting the energy found in the motion of fish to remove toxins from water (AskNature 2017). 2. Structure: BID inspires structural solutions for design problems. Structure refers to the physical orientation or placement of material (macro-scale organization). An example is a ceiling fan that mimics the structure of Sycamore seedpods to be more aerodynamic (AskNature 2002). 3. Material: BID inspires a new material to solve a design issue (micro-scale, e.g. changing molecular structure). Material includes novel chemical compounds. Examples include self-repairing concrete that mimics human bones (AskNature 2019). 4. Procedure: BID inspires a new way to perform a task. Multiple steps are required for it to be considered a procedure. A BID sample that falls into this subcategory is a visual processing software that mimics the procedure used by the human eye to identify what is being seen (Asknature 2022a). Strategy-Procedure is different than the Application-Process (but they often occur together). For example, a BID procedure (biomineralization in coral reefs) could be used to create a static object (concrete), or a BID procedure (evaporation) could inspire a new process (evaporative toilets) (Garvin et al 2015, AskNature 2020b.

Purpose of applying BID
This dimension documents the design improvement or goal of implementing BID. The initially-generated subcategories were checked against the BioTRIZ contradictions table (Bogatyrev and Bogatyreva 2014) and the UN sustainable development goals. The subcategories were not solely based on either of these sources, but they were used to support subcategory brainstorming. Additionally, each AskNature Innovation has a list of benefits for each sample, which were also checked to ensure the list of subcategories was comprehensive. Entries could be classified into more than one subcategory.
1. Reduce energy use: Applications of BID that reduce the amount of energy necessary to accomplish a task. For example, a school in Sweden was designed to mimic a termite nest to reduce the energy needed to heat or cool the school (Frangoul 2020b). This subcategory also includes innovations that improve computing efficiency or minimize purified water use due to the energyintensive nature of these processes. 2. Reduce material use: Applications of BID that reduce the amount of material necessary. Chemicals needed for agriculture or material processing are considered a material. An example in this subcategory is a new building technique that reduces the amount of building materials required by mimicking how soap bubbles and biological cells bind together (AskNature 2008). 3. Increase reliability: The application of BID to allow an object or system to be utilized for longer without interruption due to failure. This includes physical objects as well as software. Reliability is not used as a synonym for accuracy. Accuracy improvements are classified under Improve functionality. An example is a bridge design modeled on human limbs and joints which increases the flexibility of the bridge (AskNature 2020c). 4. Generate energy: Improvements in energy harvesting or generation. For example, new materials that can shift to follow the sun and gather more energy (Montalbano 2019). 5. Improve functionality: This subcategory benefits the user's experience and makes an innovation easier to utilize or more effective. For example, a camera lens inspired by the eyes of paper wasps improves image quality (Korea Advanced Institute of Science and Technology 2020). If the only improved functionality is reduced energy use (or increased reliability), then the sample is classified only under Reduce energy use (or Increase reliability).

LCA categories
The final dimension is which portion of a product's life cycle the innovation impacts. LCA is an approach to quantify a product's impact on the environment over the entire life cycle. Impact can indicate an increase or decrease in environmental health. Classifying utilizes standard LCA categories for this dimension (Finkbeiner et al 2006). Innovations are classified only into one subcategory.
1. Raw material extraction: BID applications that involve product downsizing, harvesting natural resources from the earth, or reducing the amount of material required. Chemicals are considered material. An innovation that replaces chemicals with another process would be classified here, while an innovation that changes the amount of chemicals during use is classified under the Usage subcategory. An example is biofungicide, a natural way to protect crops from fungi without using harsh chemicals (AskNature 2020a).
2. Manufacturing and processing: BID applications that affect the energy or raw materials used during product manufacturing. 3. Transportation: BID applications that support efficient distribution or reduce the environmental impact of delivery. An example is a transportation network inspired by slime molds that has equal or greater efficiency and fault tolerance than currently implemented networks (Tero et al 2010). 4. Usage: BID applications that focus on impact during use. Products that consume less power and reduce the use of auxiliary materials (e.g. water, detergent, or disposable elements) while in service are classified in this subcategory. When a product is improved so that it is easier to utilize or more effective, it is also classified in this subcategory. This subcategory also includes innovations that increase the amount of time that an innovation can be used without needing additional resources to repair or replace it. For example, a hexapedal robot is more able to climb and move around, allowing it to be utilized for a wider variety of applications (Spenko et al 2008). 5. Disposal: BID applications that make products easier to recycle or have lower amounts of environmentally harmful substances. Innovations that impact waste processing also fall into this subcategory. For example, a vortex generator inspired by how trout catch food is used to filter out harmful substances from water (AskNature 2017).

Results and discussion
Classifying 660 BID samples across seven dimensions and 68 subcategories resulted in three contributions. First, by examining the combined dataset, we obtain insights about the overall state of BID. Secondly, by identifying which subcategories are underrepresented within each source, we identify potential areas where BID could be applied more frequently. Third, by uncovering mismatches between the three sources, we provide an initial indication if there are differences between how these groups view or apply BID. Significant results are presented here, while sections 4.1-4.7 present detailed analysis, including if previous research has been done that can help explain the detected trends. The full, classified dataset is available from the researchers upon request and detailed calculations can be found in appendix B. When examining the overall results (Contribution 1), we see that BID is often used to inspire a static or dynamic physical object (54.7% of samples), with the strategy of changing the object's structure (61.4% of samples), with the goal of improve functionality (72.5% of samples), and impacts the usage phase of a product's life cycle (87.6% of samples). In the inspiration-taxonomy dimension, BID appears to be primarily inspired by commonly known sources such Significant findings for Contribution 2 (differences within each source) include the fact that only one subcategory was depressed within a source (table 5). Elevated and depressed are greater than one standard deviation from the mean within each source. The application-process subcategory was depressed in Google Scholar. This result is influenced by our approach for identifying elevated and depressed subcategories (greater than one standard deviation from the mean). Throughout the dataset, the standard deviation was often greater than the mean (16 out of 21 source-dimension pairs). The large standard deviations were because for five of the seven dimensions examined (bottom five rows in table 5), a single subcategory contained at least 35% of the classified data for that source. The only exception was AskNature-mobility classifying. Additionally, for the three dimensions eligible for χ 2 testing, no perspectives passed the χ 2 test. Finally, we observed a general agreement between the sources for which subcategories were elevated. Among the 27 entries in table 5, there are only 11 unique subcategories listed.
General observations for Contribution 3 (differences between sources) include that about half of the subcategories were consistently represented across all three sources (53% of the subcategories passed the χ 2 test) (table 6). No dimensions, however, had all subcategories pass the χ 2 test. Because these tests all had the same χ 2 test statistic, we can compare the subcategory skew across dimensions. The three most skewed (highest χ 2 values) were Applicationnetwork design (60.45), Application-static physical object (36), and Mobility-microscopic (33.06). In fact, four of the top five most skewed subcategories were in the dimension Application. This finding indicates that there are significant differences between how the three perspectives view BID application. This result is consistent with the observation that only a small portion of investigated BID (Google Scholar) has been successfully implemented (Google News or AskNature) (Lenau et al 2018). Depending on whether a subcategory was depressed or elevated, different recommendations are warranted (table 6, bottom row). Elevated and depressed are greater than one standard deviation from the mean between the sources (Contribution 3). These recommendations provide insight into possible approaches to reduce the differences between perspectives or opportunities that exist because of the differences between perspectives.
Of note, during data collection the procedure outlined in section 3.1 step 4 required modification because Google News and Google Scholar did not produce the required number of samples per equation (1). Since Google News only generates 100 results for each search term, certain terms did not provide enough BID samples. For Google Scholar, collection stopped after reviewing 500 results for each search term. The search terms with insufficient results for Google News were biomechanics, bioengineering, biological engineering, biotechnical engineering, and biophysics. The search terms that had insufficient results for Google Scholar were biomedical engineering, biomechanics, biophysics examples, biophysics, and biological engineering. Once the terms with insufficient samples were identified, the remaining samples were distributed proportionally across the remaining search terms (appendix A). This adjustment resulted in 21 samples being redistributed for Google News (9.6% of the Google News samples) and 60 samples being redistributed for Google Scholar (27.3% of the Google Scholar samples). Due to being only 12.3% of the combined planned sampling strategy, we assess that this nonideality does not negate our overall results. Additionally, by identifying Table 6. Contribution 3 summary: differences between sources and resulting recommendations elevated and depressed are greater than one standard deviation from the mean between the sources. Some recommended activities due to these results are listed in the bottom row. There is a need for practitioners and scholars to persue additional promotion and news coverage.
There are existing applications in the literature and news that could be promoted to practioners.
Opportunity for additional outreach to news agencies and practioners.
Research or application in these areas received higher than proportional news coverage.
Research in this area could receive higher than proportional use and promotion among practitioners.
which search terms did not effectively produce BID samples, we have identified certain terms that do not widely apply to BID, contrasting our and Shu et al's (2011) expectations. These terms can be used to describe innovations that utilize biomimicry, but they are not useful search terms for locating examples of BID.

Application dimension
Contribution 1 (overall): A key finding is that BID is primarily applied to physical/tangible objects (static physical objects and dynamic physical objects accounted for a combined 54.7% of the data). This result is driven by Google News (77.3% of the news samples were Physical applications). This result was expected as it is consistent with the literature. Previous design-by-analogy studies theorized that mechanical functionality is easier for engineers and designers to understand than more abstract phenomena (Lenau et al 2018, Nagel et al 2018. Additionally, BID applications are often either apparent or the result of a fortunate coincidence (Shu et al 2011). This effect was also seen in an experiment conducted by Nagel et al where 63% of their samples used physical analogies (Nagel et al 2018). Physical applications of BID are the easiest for designers to visualize (Nagel et al 2018). Results that were greater than one standard deviation from the mean sample value within/between sources are considered elevated or depressed. The complete statistical analysis for the Application dimension is found in appendix B: table B1.
All subcategories were present for all sources, with the lowest occurrence being six samples of BID applied to Software in Google News (figure 2). Contribution 2 (within sources): There were four subcategories elevated (Google Scholar-network design, Google News-static and dynamic physical object, and AskNature-static physical object) and one subcategory depressed (Google Scholar-process).
Contrasting our expectations, Network Design is the most common application in the current research (28% of Google Scholar samples). The BID samples from Google News were primarily classified into static and dynamic physical object subcategories (combined 77%). Finally, about half of AskNature's BID samples fell into the static and dynamic physical object subcategories. Of the three sources, Google Scholar examined all areas of application most evenly (lowest χ 2 test value for an underlying uniform distribution).
Contribution 3 (between sources): There were four subcategories elevated (Google Scholar-Software and Network Design, Google News-Dynamic Physical, and AskNature-Process) and two subcategories depressed (Google Scholar-Static Physical, Google News-Chemical). Across all subcategories, only the Process subcategory passed the χ 2 test for underlying uniform distribution between perspectives.
The elevation occurrence of network design from the academic perspective is consistent with observations that novice designers may have difficulty with 'system-level' problems because of their abstractness (Weissburg et al 2010, Yen andWeissburg 2012). This point is not to say that the public or practitioners are novices, rather that these results seem to indicate that a narrower focus (as seen in academic research) supports the application of BID to The 13 smallest subcategories are grouped together as 'other' to ensure readability. Full results are in appendix B: table B2. For Contributions 2 and 3, the blue dashed lines are the mean sample value within/between sources, while the blue box shows one standard deviation from the mean sample values. Results that were greater than one standard deviation from the mean sample value within/between sources are considered elevated or depressed. network design. The higher occurrence of physical applications in Google News and AskNature supports the observation that practical applications are often focused on manufactured goods rather than on network studies (Chatterjee and Layton 2019). Our suggestion to reduce this disparity between perspectives is to focus on tools to translate academic findings into practical applications. This recommendation mirrors suggestions in the literature (Layton et al 2016).

Inspiration taxonomy
Contribution 1 (overall): For the inspirationtaxonomy dimension there are 26 subcategories. The subcategory with the highest occurance across all sources is Mammals at 27.8% ( figure 3). The next highest percentage (17.8%) was in the Arthropods subcategory. BID samples in this dimension could be classified into more than one subcategory, so the percentages are from the total number of classified results (701) and not the total number of samples (660).
The prevalence of mammals in the dataset is not surprising as it could be a manifestation of bias toward 'charismatic megafauna.' 'Charismatic megafauna' enjoy higher levels of support than other species, because of their easier emotional connection with the population (Kollmuss and Agyeman 2002). Arthropods could also be elevated as a practical result of lower institutional review board oversight requirements for arthropod research compared to other members of Kingdom Animalia. The focus in these two areas also reflects that BID is often focused on familiar inspiration sources (Mak and Shu 2004, Shu et al 2011, Linsey and Viswanathan 2014. This density could also reflect case-based reasoning, where previous sources of inspiration were used to solve new design challenges (Goel 1997). Surprisingly, the only subcategory that did not have a sample was Amphibian.
Overall, about half of the subcategories combined accounted for less than 1% of the samples. Exposing designers to these less frequent sources of inspiration could reduce design fixation (Linsey and Viswanathan 2014), but could be more cognitively difficult due to being unfamiliar (Nagel et al 2018).
Our findings indicate that the recommended practice of considering a wide range of inspirations (Nagel et al 2018) could be implemented more. Of note, Nagel points out that the wide range of inspiration needed could come from varying species (as recorded in this dimension) or varying levels of detail of the same species. The second approach is expected to be less cognitively difficult for designers (Nagel et al 2018). We recommend exploring a wide range of inspirations to enable the identification of common solutions across different species, indicating solution robustness (e.g. single lens eyes in squid and humans). Exploring different related species is also recommended in order to refine our understanding of potential biological solutions. For example, Casper and Müller investigated bat sonar from 98 different species to identify the elements required for effectiveness (Caspers and Müller 2015).
Contribution 2 (within sources): Each source had at least five subcategories with no samples classified into them. Google Scholar had an elevated representation of both Mammal and Arthropod inspiration (a combined 50.6%). Google News and AskNature had elevated representation of Mammal, Plant, and Arthropods. While this analysis assumes an underlying uniform distribution of taxonomy on the earth available for inspiration for the three groups (an unrealistic assumption), the identification of elevated and depressed subcategories still provide value, because, for example, we know that 27.8% of species on earth are not mammals, despite 27.8% of the current samples being mammals. The χ 2 test within perspectives is not applicable.
Contribution 3 (between sources): For Google Scholar, six subcategories were depressed (Reptile, Fish, Plant, Mollusk, Nematode, and Cnidarias) while three were elevated (Cell, Enzyme, and Molecule). For Google News, five were depressed (Slime Mold, Amoeba, Algae, Bacteria, and Tardigrade) while four were elevated (Annelid, Platyhelminth, Echinoderm, and Poriferous). Finally, for AskNature, one was depressed (Mammal) and four were elevated (Arthropod, Plankton, Ecosystem, and Natural phenomena). Of note, Mammal is overrepresented within AskNature, but less present among practitioners than for academics or the public.
These findings could be the result of the conceptual leap hypothesis: more distant analogies have lower efficacy in analogical transfer (lower occurrence in Google Scholar) but higher potential for revolutionary results (higher visibility in Google News and AskNature) (Lenau et al 2018, Jiang et al 2022. We recommend additional fundamental research into BID inspired by Reptile, Fish, Plant, Mollusk, Nematode, and Cnidarias. This research may have a higher chance of breakthrough results than more familiar sources of inspiration. The distant analogies that are more popular for Google Scholar are in microscopic subcategories (Cells, Enzymes, Molecules). This higher prevalence is potentially due to the interest in biomedical research.

Inspiration-biome
Contribution 1 (overall): Although the Biome dimension has 13 subcategories, 61.3% of the samples were Worldwide/unknown (figure 4). While most sources referred to where they gained their inspiration from, they did not provide sufficient details to determine the biome of the source. For example, some sources would state that their innovation was inspired by an insect, but they did not specify the type of insect or where the insect was from. Thus, we recommend detailing as specifically as possible the source of inspiration in order to aid future researchers. Additionally, if the source of inspiration was found in more than three biomes, then it was classified into the Worldwide/unknown subcategory. Having multiple sources of inspiration is not uncommon in BID; one previous study found that 66% of the solutions examined were composed of multiple analogies . The second most frequent subcategory was Marine (15.9%). Overall, over 20% of the BID samples fell into a water subcategory (Marine or Freshwater), highlighting the importance of aquatic inspirations. BID samples in this dimension could be classified into more than one subcategory, so these percentages are taken from the total number of classified results (734) and not the total number of samples (660). These findings reveal that more systemic and intentional exploration of different biomes for inspiration could be a fruitful avenue for additional research.
Contribution 2 (within sources): Consistent with the overall observations, all three sources showed overrepresentation for the same subcategory: Worldwide/unknown. There were no other results within each source that were more than one standard deviation from the mean per subcategory for that source. The χ 2 test is not applicable. Google Scholar had the most skewed distribution, with the highest standard deviation and 81.7% of the samples being classified as Worldwide or Marine. The four smallest subcategories are grouped together as 'Other' to ensure readability (arctic tundra, mountains, temperate grassland, and tropical seasonal forest). The statistical analysis for the inspiration-biome dimension is found in appendix B: table B3. For contributions 2 and 3 the blue dashed lines are the mean sample value within/between sources while the blue box shows one standard deviation from the mean sample values. Results that were greater than one standard deviation from the mean sample value within/between sources are considered elevated or depressed.
Contribution 3 (between sources): For Google Scholar, one subcategory was elevated (Worldwide/unknown) and five subcategories were depressed (Marine, Temperate forest, Temperate grassland, Desert, and Tropical savanna grassland) relative to the other sources. For Google News, the Northern coniferous forest and Mediterranean vegetation subcategories were both elevated. Finally, for AskNature, Tropical seasonal forest was elevated. For recommendations due to these differences, see table 6.

Inspiration mobility
Contribution 1 (overall): The inspiration mobility dimension was difficult to classify because many samples only included minimal information about their sources of inspiration. For example, some sources refer to their source of inspiration as insects, but insects can fly, swim, climb, and walk. Thus, we recommend detailing as specifically as possible the source of inspiration in order to aid future researchers. BID examples in this dimension were able to be classified into more than one subcategory, so the percentages are taken from the total number of classified results (807) and not the total number of samples (660). Overall, the walking subcategory had the highest percentage of BID samples (37.3%), but this was expected because most animals can walk and then climb or fly as well. The next highest occurrence was 14.7% in the static subcategory.
Contribution 2 (within sources): For all three sources, walking was the only elevated subcategory because most sources of inspiration can walk in Figure 5. Inspiration-mobility results. The statistical analysis for the inspiration-taxonomy dimension is found in appendix B: table B4. For contributions 2 and 3, the blue dashed lines are the mean sample value within/between sources while the blue box shows one standard deviation from the mean sample values. Results that were greater than one standard deviation from the mean sample value within/between sources are considered elevated or depressed.
conjunction with other forms of mobility (figure 5). AskNature showed the most even distribution within a source (lowest standard deviation).
Contribution 3 (between sources): The swimming, flying, and static subcategories are depressed, but microscopic motion is elevated for Google Scholar. This result is unsurprising considering that the subcategories of cells, enzymes, and molecules in the taxonomy dimension were also elevated for Google Scholar. Google News has no subcategories that vary from the average by more than the standard deviation. For AskNature the walking subcategory is depressed, and the crawling, climbing, and N/A subcategories are elevated relative to the other sources, but this resulted in AskNature having a more even distribution (contribution 2). For recommendations due to these differences, see table 6.

Strategy dimension
Contribution 1 (overall): For this dimension, most samples were in the structure subcategory (61.4%). This result may reflect the use of bioreplication as an approach to BID. Bioreplication is a solution-driven approach that mimics the structures or forms found in nature (e.g. kingfisher's beak shape applied to train noses) (Lenau et al 2018). Vincent at al noted that when comparing engineering and biological problem solving, the approach to mimic spatial orientation is the most common (Vincent et al 2006). Thus, structure transfer might have less friction due to using approaches familiar to engineers. Therefore, it might be advisable to use structure transfers when teaching engineers new BID approaches.
The next most prevalent strategy was procedure (20%) followed by material (15.8%). Material, being Figure 6. Strategy dimension results. The statistical analysis for the strategy dimension is found in appendix B: table B5. For contributions 2 and 3, the blue dashed lines are the mean sample value within/between sources while the blue box shows one standard deviation from the mean sample values. Results that were greater than one standard deviation from the mean sample value within/between sources are considered elevated or depressed. second most uncommon, is consistent with a previous experiment with graduate students that found that material was unlikely to be transferred (Nagel et al 2018). Biological materials are made with much fewer components than are available to engineers, are often anisotropic (different properties when measured across different dimensions), frequently selfassemble, are hierarchical, and are multifunctional (Vincent et al 2006, Bhushan 2009, Meyers et al 2011, Pan 2014. Of note, there is a recognized research opportunity to explore the hierarchical structure of biological materials that still appears to be available (Meyers et al 2011, Pan 2014. One significant challenge in this area is transferring biological inspiration across multiple scales and strategies (Bhushan 2009, Pan 2014. Using different natural material properties at different scales allows improved performance, but multi-scale material design is often difficult due to disciplinary boundaries (material science versus mechanical or structural engineers) (Pan 2014). In nature, material properties are often associated with specific structural orientations (Bhushan 2009, Meyers et al 2011, Pan 2014, therefore if designers attempt to transfer material and structure separately, this could severely limit their ability to utilize bioinspired materials. We recommend also considering structure when investigating material inspirations.
The use of energy to solve design problems only received 2.9% of the BID samples, which is consistent with previous research that found that biological inspirations rarely used energy or new materials to solve challenges (Vincent et al 2006, Shu et al 2011. Contribution 2 (within sources): For all three sources, only the structure subcategory was elevated (figure 6). Within each source, all failed the χ 2 test for underlying uniform distribution within the source. Purpose dimension results. The statistical analysis for the strategy dimension is found in appendix B: table B6. For contributions 2 and 3, the blue dashed lines are the mean sample value within/between sources while the blue box shows one standard deviation from the mean sample values. Results that were greater than one standard deviation from the mean sample value within/between sources are considered elevated or depressed. This result indicates that the strategy used to apply BID is not evenly spread among the four types. There is an opportunity for our three perspectives to investigate additional approaches to using BID to impact energy use to solve engineering challenges.
Contribution 3 (between sources): Google Scholar-Energy was slightly depressed, and Google Scholar-Procedure was elevated. For Google News, only the structure subcategory was slightly elevated. Finally, for AskNature, only the material subcategory was slightly elevated. All the BID sources displayed similar distributions of samples, but energy and procedure passed the χ 2 test for underlying uniform distribution between sources. Surprisingly, each source had a different overrepresented subcategory (procedure for Google Scholar, structure for Google News, and material for AskNature). For recommendations due to these differences, see table 6.

Purpose dimension
Contribution 1 (overall): 72.5% of the data collected fell into the improve functionality subcategory (figure 7). The next highest percentages in this dimension were 12.8% in the increase reliability subcategory and 11.2% in the reduce energy use subcategory. The reduce material use and generate energy subcategories had the smallest presence (2.1% and 1.4%), respectively. Because BID samples in this dimension can be classified into more than one subcategory, the reported percentages of the total number of classified results (805) and not the total number of samples (660).
Contribution 2 (within sources): For all three sources, the improve functionality subcategory was elevated. This finding is consistent with previous observations that many BID solutions may not be more environmentally sustainable because of the manufacturing complexity of the products or  : table B7. For contributions 2 and 3, the blue dashed lines are the mean sample value within/between sources while the blue box shows one standard deviation from the mean sample values. Results that were greater than one standard deviation from the mean sample value within/between sources are considered elevated or depressed. chemicals involved (Lenau et al 2018). We expect that the improve functionality subcategory would have examples of improved performance but not improved sustainability, while reduce energy use, reduce material use, or increase reliability would more closely parallel environmental impact. The χ 2 test within perspectives is not applicable.
Contribution 3 (between sources): For Google Scholar, the reduce energy use subcategory was depressed. For Google News, the generate energy subcategory is slightly depressed, but only failed our testing criteria by 0.1. Finally, for AskNature, the reduce material use and increase reliability subcategories were elevated, and the improve functionality subcategory was depressed. Although AskNature-improve functionality was elevated within the public perspective, it was depressed relative to the other perspectives.
For recommendations due to these differences, see table 6.

LCA dimension
Contribution 1 (overall): These results indicate that BID primarily impacts the use phase (87.6%). The next highest occurrence was for raw material extraction (5.6%). A previous study noted that much of the BID research in the manufacturing domain was motivated by sustainable manufacturing processes, but, surprisingly, these results indicate that BID is applied more often during the use phase (Shu et al 2011).
Contribution 2 (within sources): Once again all three sources agreed and indicated that BID application is focused on the use phase (figure 8). Interestingly, Google News had no samples in manufacturing or transportation. None of the sources passed the χ 2 test, but AskNature did have the lowest test value, indicating a more even distribution within that source. Given the current emphasis on the circular economy, there appears to be an opportunity for BID applications in disposal or manufacturing stages.
Contribution 3 (between sources): For Google Scholar and Google News, none of the subcategories were elevated or depressed. AskNature, however, had an elevated representation in raw material extraction, manufacturing, and disposal, as well as being underrepresented in usage. This finding is not necessarily a negative observation; it could indicate a strategy by AskNature curators to display innovations across the product life cycle. Transportation and usage subcategories both passed the χ 2 test between sources, indicating that these subcategories have equal representation across all three sources.

Conclusion
This study gathered and evaluated a dataset of 660 samples of BID to examine the research question: 'how does the purpose of applying, the inspiration source, and the application of BID differ between academics, the public, and practitioners?' This investigation makes three contributions. First, we provide insight into the overall state of BID, regardless of perspective. BID is most often used in the physical domain, inspired by a part of nature that is commonly well-known, uses a structural strategy to solve an issue, improves an innovation's reliability, and impacts the use phase. Second, the analysis reveals focus areas within each of the three perspectives. For example, in four of the dimensions, a single subcategory had more than 60% of the data (biomeworldwide/unknown, strategy-structure, purposeimprove functionality, LCA-usage). Finally, the investigation revealed mismatches between the three sources (table 6). For example, in the purpose dimension, Google Scholar-reduce energy use, Google News-generate energy, and AskNature-improve functionality were depressed.
Considering previous research, the researchers were expecting most of the results of contribution 1. They were expecting BID to be primarily utilized in the physical domain with structural strategies to improve the reliability and usage phase of a product. For contribution 2, the researchers were not expecting there to be such a clear focus in most dimensions on one specific subcategory. Finally, for contribution 3, the researchers did not expect there to be significant mismatches in the sources. However, each source proved to have a slightly different focus for how to apply BID. These unexpected results further illustrate the importance of studies like this one that evaluate the current state of practice.
This study provides insight into the current application of BID across multiple areas, but does not provide an exhaustive analysis of the state of BID. One limitation is that this study was unable to identify examples of BID that were not described using one of the search terms in table 2. For instance, if a company created a product inspired by the properties of a plant and did not state it was a type of BID, then this study would not locate the example. There is also room for additional studies that seek to understand the 'state of BID' . This paper uses two Google platforms for collecting data. Future investigations could use other data sources to verify our results. It would be particularly helpful to use a single platform or data source for the public perspective (rather than combining Google News and Google Trends). Future work could also include explicitly examining causality between the sources. For example, does a Google Scholar publication cause AskNature and/or Google News, or does Google News promote additional work in an area, leading to additional scholarly investigation. Future work could also focus on testing whether the type of BID tool (e.g. BioTRIZ) impacts the characteristics of the BID application. Additionally, future work will investigate how to close the gaps in the differences between perspectives. This paper focuses on identifying these differences, but there is room to determine detailed approaches for resolving these differences. Similarly, the differences between subcategories are highlighted, but the reasons for the differences are beyond the scope of this paper. Future research will investigate the causes behind the variations between the subcategories, and if the main reason for the differences is due to the difficulty of applying BID in specific areas.
This initial examination of BID has created a resource for examining how BID is applied and viewed. These findings support future work that focuses on developing novel applications of BID and better tools to support BID practitioners. Our goal is to use BID to improve modern innovations and shed light on the potential of BID for new areas of research.

Data availability statement
All data that support the findings of this study are included within the article (and any supplementary files).    Contribution 2: within sources (↑, ↓) Table B3. Inspiration-biome dimension analysis results for any contribution that is suppressed are italicized, results that are elevated are bold. Elevated and depressed are greater than one standard deviation for the mean (SAM) within or between the sources. For contribution 2 (within sources) ↑ indicates elevated and ↓ indicates depressed. For contribution 3 (between sources), ⊕ indicates elevated and ⊖ indicates depressed. To determine if there is even representation between sources, a χ 2 test was performed (for an underlying uniform distribution between sources). For χ 2 results, light gray shading indicates a failed χ 2 test, while dark green shading indicates a passed χ 2 test.

Sources
Contribution 1: overall  Table B4. Inspiration mobility dimension analysis results for any contribution that is suppressed are italicized, results that are elevated are bold. Elevated and depressed are greater than one standard deviation for the mean (SAM) within or between the sources. For contribution 2 (within sources) ↑ indicates elevated and ↓ indicates depressed. For contribution 3 (between sources), ⊕ indicates elevated and ⊖ indicates depressed. To determine if there is even representation between sources, a χ 2 test was performed (for an underlying uniform distribution between sources). For χ 2 results, light gray shading indicates a failed χ 2 test, while dark green shading indicates a passed χ 2 test.

Sources
Contribution 1: overall  Table B5. Strategy dimension analysis results for any contribution that is suppressed are italicized, results that are elevated are bold. Elevated and depressed are greater than one standard deviation for the mean (SAM) within or between the sources. For contribution 2 (within sources) ↑ indicates elevated and ↓ indicates depressed. For contribution 3 (between sources), ⊕ indicates elevated and ⊖ indicates depressed. To determine if there is even representation between or within sources, a χ 2 test was performed (for an underlying uniform distribution). For χ 2 results, light gray shading indicates a failed χ 2 test, while dark green shading indicates a passed χ 2 test.

Sources
Contribution 1: overall   Table B7. Life cycle assessment stage dimension analysis results for any contribution that is suppressed are italicized, results that are elevated are bold. Elevated and depressed are greater than one standard deviation for the mean (SAM) within or between the sources. For contribution 2 (within sources) ↑ indicates elevated and ↓ indicates depressed. For contribution 3 (between sources), ⊕ indicates elevated and ⊖ indicates depressed. To determine if there is even representation between or within sources, a χ 2 test was performed (for an underlying uniform distribution). For χ 2 results, light gray shading indicates a failed χ 2 test, while dark green shading indicates a passed χ 2 test. ORCID iD B C Watson  https://orcid.org/0000-0003-2222-6716