Ways forward for global adaptation evidence synthesis building on the Global Adaptation Mapping Initiative

Abstract

Global evidence of climate change adaptation is increasing faster than assessment capabilities of conventional systematic review approaches. To inform better decisions based on this growing literature, the Global Adaptation Mapping Initiative deployed AI-assisted and crowd-sourced methods for synthesising and tracking adaptation globally, offering important knowledge for the Intergovernmental Panel on Climate Change and the Global Stocktake. Here we present results of a survey with Global Adaptation Mapping Initiative researchers regarding the lessons learned and recommendations for future global evidence syntheses on adaptation: transparently reflect on the compromises regarding breadth vs depth of the synthesis; employ the highest standards of quality assurance for human coding; promote justice and equity within synthesis teams; extend the scope beyond academic literature; and further integrate machine learning methods. Building on these learnings, future work can contribute to a more holistic understanding of adaptation progress and raise the standard of adaptation evidence synthesis.

Introduction

The increasing impacts of climate change have led to a general understanding that mitigating further global warming alone is insufficient; there is an urgent need for adaptation action that reduces current and future climate risks¹. Evidence of adaptation taking place in various systems and sectors around the globe is growing^2,3. However, this evidence base is highly diverse regarding geographical coverage and methods to study how adaptation is happening, by whom, and how (and whether and when) it reduces climate risk. Therefore, evidence synthesis methods have become increasingly relevant to understanding where the world stands regarding adaptation progress^4,5.

Such evidence synthesis in the field of climate change adaptation is of high political and practical relevance across scales: from local-level adaptation to international arenas such as the Global Stocktake and the Global Goal on Adaptation. Reliable information about what adaptation approaches work, to what degree, and under what conditions requires knowledge regarding which adaptation is being implemented, the progress on adaptation in different sectors and regions, and where science stands in evaluating (new) adaptation options. While the growing body of evidence on adaptation practices allows the research community to undertake systematic evidence syntheses⁶, it is not yet a common practice, unlike in other scientific fields such as medicine and environmental science.

The Global Adaptation Mapping Initiative (GAMI) was, to the best of our knowledge, the first network-based systematic evidence synthesis project that mapped global adaptation literature², the outputs of which informed the Intergovernmental Panel on Climate Change (IPCC) 6th Assessment Report (AR6) of Working Group II. GAMI was a unique, scientific community-driven process, with 129 researchers globally. The team used machine learning to screen over 40,000 studies and in-depth manual coding to analyse over 1600 articles to create a comprehensive database. The original GAMI publication quickly became a key reference point for assessing the global state of human adaptation. Eighteen additional studies have been published, using the GAMI data to synthesise specific sub-questions including many adaptation themes that had lacked global assessment, for example, on adaptation in critical sectors including water⁷ and health⁸, the evidence on adaptation limits across sectors and regions⁹, different policy tools supporting adaptation¹⁰, the roles of different actors in adaptation¹¹, adaptation in mountains¹², forests¹³, coastal cities¹⁴, adaptation feasibility assessment¹⁵, adaptation to compound climate risks¹⁶, and the interaction between autonomous and planned adaptation¹⁷.

Now it is critical to take stock and examine the lessons learned for the future of GAMI vis-à-vis evidence synthesis in climate change adaptation. The study of team science has identified many of the structural and interpersonal challenges that shape large-scale collaborations¹⁸. Yet the field continues to call for more empirical data on how scientific teams actually experience collaboration, particularly through systematic self-reflection by contributors working within large, distributed teams—an area where existing evaluations remain limited^19,20. Here we use a structured self-reflective survey of global participants to examine what worked well in the GAMI process, the challenges and limitations, and—most importantly—the opportunities and recommendations for urgently needed global adaptation evidence synthesis in the future. Specifically, what are the potential contributions of multi-method global evidence synthesis to future scientific and policy assessments of adaptation?

We present the results of a survey collecting the experiences of researchers involved in GAMI in different roles (see Methods) and their suggestions moving forward. We collected 59 completed surveys (46% of 129 invited GAMI researchers), including 12 of the 17 (71%) corresponding authors of papers that used the GAMI database for specific spin-off analyses, who shared their unique insights about the opportunities and limitations of the dataset, extensions and re-analyses of the dataset, learnings for future global adaptation mapping efforts, and integration with novel machine learning-assisted methods. Results are presented around four main criteria that apply to adaptation evaluations: the impact of the results in science, policy, and media; the coverage of the GAMI approach and data; its reliability, validity, and usability; and its feasibility and efficiency.

The synthesised lessons learned from the GAMI process can lay the basis for improved adaptation evidence synthesis methods, which are crucial for a robust science-informed Global Stocktake, assessing progress towards the Global Goal on Adaptation, and environmental assessments, such as GEO, IPBES, and in particular the 7^th IPCC assessment report.

Results

Impact of the results within science, policy and media

Adaptation evidence syntheses aim to improve our scientific understanding of adaptation while offering knowledge for adaptation decision-making. Therefore, they must be evaluated against their impact on the target audiences. For the case of GAMI, audiences include academia, policy and practice as well as media uptake (Fig. 1). The perceived impact of GAMI was assessed through a self-report question asking researchers to rate the visibility and influence of GAMI findings in research, media, and policy contexts.

Fig. 1: Reception of GAMI across research, media, and policy arenas, as reported by GAMI researchers.

Responses (n = 55) were based on researchers’ subjective assessments of the visibility and influence of GAMI findings in each domain.

Respondents report that GAMI and its results have been received positively. According to Nature Climate Change Altmetrics (as of 16 January 2026), the first GAMI paper was viewed 34,000 times and cited 522 times, and various spin-off papers received similar academic attention. How academic peers have picked up GAMI can also be understood through the feedback in the peer reviews of GAMI spin-off publications, which we asked GAMI authors to reflect on. GAMI papers received positive feedback from reviewers, highlighting the comprehensive overview across a vast body of literature, the database’s global scope, and large sample size. However, reviewers also provided more critical feedback, including questions about how far the synthesis of empirical studies reflects real-world adaptation. Additionally, some reviewers raised concerns regarding the high level of aggregation of case study data, geographic bias, and outdated data, since the GAMI dataset included papers published only through 2019.

In terms of the uptake in the policy space, GAMI researchers felt that their results have been received positively, for example, among UNFCCC audiences. GAMI was cited, amongst others, in reports by the World Bank²¹, UN-HABITAT²², UNEP²³, FAO²⁴, Overseas Development Institute²⁵, and the OECD²⁶. Moreover, respondents highlight that the findings from GAMI informed high-level statements on adaptation in the IPCC AR6 Working Group II Report’s Summary for Policymakers²⁷ and Synthesis Report²⁸.

Several involved GAMI researchers were unaware of or rated critically the reception of the media, noting a lack of media attention focused explicitly on GAMI’s results, calling for “more coverage in the future”. However, some interest has been generated due to the global nature of GAMI, and a few research institutions contributed to the promotion of the research. According to Nature Climate Change Altmetrics (as of 16 January 2026), the primary GAMI paper appeared in 40 news stories, 14 policy sources, and 308 Twitter (now X) posts from a dozen countries, with the majority of the attention in the first six months after publication. An attention score of 480 places the paper in the 95th percentile of articles of a similar age in the same journal.

Beyond the direct impact and reception of the GAMI results in research, media, and policy, a major impact of the initiative was the further community-building of researchers from diverse disciplinary backgrounds on global adaptation evidence synthesis. For example, one European respondent mentions that “collaborations amongst the GAMI members have strengthened with members trying to collaborate [on new efforts]”, and a respondent from Africa explains that GAMI provided “opportunities for collaboration and networking, especially for early career researchers from global south”. The survey responses and feedback from the authors of spin-off articles reflect a high motivation to “regroup and plan for GAMI 2.0 with the lessons learnt from GAMI 1.0”.

Coverage of the GAMI approach and data

Global assessments aim for a comprehensive coverage of available literature. The GAMI approach focused explicitly on scientific literature on observed human adaptation through a systematic search with English keywords (but including non-English literature) indexed in three databases²⁹. This approach, therefore, underrepresents non-English scientific literature, grey literature or other forms of evidence. This was a deliberate compromise in the GAMI design³⁰; hence, GAMI papers carefully referred to the evidence reported in the peer-reviewed scientific literature. However, the question has always been how far this evidence reported in the scientific literature represents adaptation on the ground, or whether and to what extent mismatches exist from under- or overreporting in terms of geography, sectors, informality, or other facets of adaptation.

The collected experiences show mixed views on whether the scope of what is considered ‘climate change adaptation’ should be broadened (Fig. 2). On the one hand, a broadening is considered essential to account for activities that may not be explicitly labelled as adaptation and to acknowledge that more transformative approaches to adaptation go beyond a narrow interpretation of reducing risks. On the other hand, concerns exist regarding the potential conflation of issues arising from such broadening, especially regarding the existing challenges in attributing specific interventions to reducing climate risk, which is even more difficult for generic development efforts, such as poverty reduction. For example, a respondent states: “The term ‘adaptation’ risks losing meaning if the scope is broadened too much, although I am sympathetic to arguments that other types of activities should be included or that it should be conceptualised more broadly beyond purposeful responses to climate change impacts. A challenge of any research in this area is defining adaptation clearly.”

Fig. 2: Broadening the scope of included literature.

Evaluation of the usefulness of broadening the scope of adaptation-related literature.

Since GAMI only includes peer-reviewed scientific literature, there is potential to widen the scope of literature types in future global synthesis efforts. There is very high agreement that government reports, policy documents, and reports from non-governmental organisations are important (Fig. 3). Moreover, reports from multilateral governance institutions, statistical data, legal documents, risk and adaptation indices, dissertations, and working papers are important. On the other hand, conference proceedings and preprints are considered relatively less important, mainly because they are usually not peer-reviewed. Other suggestions include reports from companies, think tanks, Indigenous organisations, planning documents, traditional and social media, and laws and regulations.

Fig. 3: Sources beyond peer-reviewed literature.

Evaluation of the importance of other sources beyond peer-reviewed literature.

Reliability, validity, and usability

The reliability (in how far the coding is accurate), validity (in how far the method allows to measure what it is supposed to measure) and usability (in how far the output can be used for further research and decision making) of globally aggregated adaptation knowledge rely on the methods of synthesis, aggregation and the accuracy of the human coding. The GAMI survey respondents perceived the overall reliability, validity, and usability of the GAMI database as rather positive (Fig. 4), reflecting the several steps to ensure quality and assess potential bias of the synthesis in the GAMI process, including training, double-coding, consistency checks and a pilot expert elicitation³¹. However, GAMI researchers note potential for improvement in the data extraction and synthesis processes.

Fig. 4: Evaluation of the GAMI database.

Evaluation of reliability, validity, and usability of the GAMI database.

The perceived reliability of the coding reported by GAMI researchers is challenged, amongst others, by the global scope, the diversity of disciplines covered and their accompanying epistemological commitments, and the highly diverse categories of interest. GAMI covered different geographic and sectoral contexts and concepts such as adaptation limits, evidence of risk reduction, and depth, scope, and speed of interventions. The standardised GAMI codebook was hence designed to cover a very heterogeneous body of literature, including empirical studies employing diverse methods, study designs, and epistemological approaches. A respondent summarises that “although the coding rubric and process were well-defined, individual coders’ interpretations of specific questions, key concepts, and codes varied” and highlights coders’ diversity in “disciplinary knowledge and perspectives to their work” and the “complexity of having about 100 coders working together on a large dataset”. In addition, experiences by GAMI researchers show, amongst others, that aggregating data from multi-region case study publications or synthesising findings from in-depth case studies to match the standardised codebook can result in the loss of context and case study-specific nuance.

The validity of the GAMI database refers to how far its synthesis reflects the current scientific knowledge of adaptation. Due to its explicit focus on adaptation-related research, the GAMI database covers a broad range of relevant literature. However, the database potentially underrepresents literature on efforts in regions and communities using different terminologies from the mainstream adaptation community (e.g., development studies). Experiences from sector-specific spin-off papers, for example, on coastal cities, showed that broader search terms may provide further relevant evidence¹⁴. Moreover, while GAMI explicitly focuses on peer-reviewed literature, it does not include scientific evidence not indexed in the databases searched. Hence, one respondent notes that GAMI is “lacking evidence … that may indirectly/implicitly capture climate change adaptation”.

Survey respondents evaluated the usability of the GAMI database as high, with some intrinsic limitations due to its global scope. Critical respondents noted that data aggregation from studies covering multiple countries hampered attributing results to specific locations. One respondent mentions that “broad trends captured by the GAMI dataset are accurate, but there’s not enough detail for deeper analysis”. Therefore, further analyses, for example, concerning specific regions (e.g., cities, islands), or topics (actors and roles, equity and justice), required additional coding or re-coding of specific categories. Experiences by non-GAMI researchers or non-academic users are not available.

Feasibility, efficiency, and equity

In general, assessments based primarily on synthesising global-scale literature through manual coding and review require considerable participation by a large group of researchers, their time commitment, topic expertise, coordination, and efficient workflows. GAMI included a coordinating and advisory team responsible for setting up the assessment infrastructure and training volunteers to screen and code literature. The coordinating and advisory teams were comprised predominantly (72%) of individuals from the Global North, representing an imbalance in the project’s leadership (see Methods for further breakdown of team membership). To support more equitable access to opportunities to use the GAMI database, the team established explicit authorship standards and a transparent process for proposing, leading and participating in spin-off analyses. Six of the 17 lead authors on spin-off analyses were from LMICs (see “Methods”). GAMI also provided clear opportunities for participants to take on larger roles, so teams established later in the process had better representation (e.g., 40% of the synthesis team were from LMICs).

Most respondents found none of their respective tasks very or extremely difficult (Fig. 5). A third of the respondents found tasks related to extracting data during the coding and resolving coding conflicts moderately difficult. In addition, committing time was seen as the most difficult aspect overall. Time constraints were more prominent among respondents from the Global South. One Global South respondent, for example, said “The amount of workload and time given” was a challenge and “much of the process was done during COVID lockdown, which had a load of other challenges, so my state of mind and physical health were probably stretched so much that this GAMI process was probably more challenging than it could have been”. An early career researcher also had difficulties with the workload as the GAMI effort required “creating time - as I submitted my PhD thesis around the same time.” But several senior researchers also found it “difficult to commit much time to unfunded work when IPCC is already unfunded”.

Fig. 5: GAMI work steps.

Perceived difficulties in GAMI work steps.

Interestingly, understanding the codebook and training, using the Sysrev software, and communicating within the GAMI team were considered the least difficult tasks overall. This suggests that most of the materials and support offered were relatively well organised.

For identifying eligible articles, GAMI built on manual human and integrated machine-learning-assisted screening²⁹. However, the whole coding process of 1,682 papers was done manually by the coding teams. To support the coding and analysis process, many GAMI researchers find machine learning-assisted approaches, such as Natural Language Processing (NLP), useful (Fig. 6). Perceived opportunities from NLP include the capacity to handle larger sample sizes, detect patterns at scales not visible to an individual reader, and facilitate the creation of living databases, which can be continually updated. However, integrating NLP into evidence synthesis methodologies also presents challenges, for example, concerning the accuracy and reliability when considering varying forms and sources of evidence. Also, the survey shows that many researchers are unfamiliar with these approaches. To maximise their reliability, respondents consider it vital to combine NLP with other analytical methods, ensuring a comprehensive understanding of the adaptation landscape.

Fig. 6: Relevance of Natural Language Processing.

Perceived relevance of Natural Language Processing (NLP) for adaptation evidence synthesis.

Discussion

GAMI represents a landmark effort in synthesising global evidence on human adaptation to climate change. It demonstrates the power of large-scale synthesis to reveal macro-level trends, such as the predominance of fragmented and incremental adaptation². However, the critical feedback from over 50 researchers involved in the initiative highlights how the methodological and epistemological choices embedded in GAMI shaped the knowledge that emerged from this effort. While offering valuable insights into adaptation patterns and knowledge gaps, the substantive findings on adaptation globally were influenced by the structural limitations and trade-offs inherent in the GAMI approach. This critical self-reflection of the GAMI community revealed key concerns that are relevant for a more comprehensive understanding of the current knowledge on global climate change adaptation progress, and the general methodological advancement of robust evidence synthesis standards.

A central theme across the GAMI-based literature that is also reflected in this survey is the tension between breadth and depth. Reflections from the GAMI community indicate GAMI’s breadth came at the cost of contextual nuance. The standardisation required for coding across diverse geographies and sectors compromised detail on case-specific insights, particularly noted in studies like the review of adaptation in forests¹³ or coastal cities¹⁴, where local socio-ecological dynamics are critical to understanding adaptation outcomes.

The aggregation of data from multi-region or multi-case studies can dilute the specificity of adaptation processes. This is particularly problematic in contexts like small islands³², where adaptation is deeply embedded in local histories, cultures, and governance structures. The resulting synthesis may inadvertently flatten these complexities, presenting a homogenised view that underrepresents the diversity of island realities. Future GAMI cycles could adopt a hybrid approach—maintaining a core standardised framework for cross-regional and cross-sectoral comparison while allowing for context-specific modules to capture local nuances, such as Indigenous knowledge systems, informal governance, or place-based adaptation practices. This may also help address underlying epistemological differences prioritising generalisability over contextual meaning. In addition, future GAMI work could actively integrate Indigenous and local knowledge systems through co-design with knowledge holders, using culturally appropriate methods and ensuring equitable recognition and ownership of knowledge. This includes developing protocols for ethical engagement and data sovereignty.

The methodological choices underpinning GAMI introduced a notable academic and linguistic bias associated with reliance on peer-reviewed literature indexed in English-language databases. This is evident in the study on water scarcity in Africa⁷ and the feasibility assessment of African adaptation options¹⁵, both of which underscore the underrepresentation of local, informal, and non-English sources of knowledge. These omissions are not merely technical oversights; they shape the contours of what is visible as ‘adaptation’ in the global evidence base, potentially skewing our understanding of where and how adaptation is occurring.

The human coding process itself, while supported by rigorous training and quality assurance protocols, also presents epistemic challenges to assessment. The diversity of disciplinary backgrounds among coders, combined with the complexity of adaptation contexts and concepts^16,17, challenges the reliability and validity of the synthesised data. The global assessment of adaptation actors and their roles¹¹ illustrates this well: while it identifies dominant institutional actors, the coding framework may obscure more informal or emergent forms of agency that are less legible within standardised categories.

The exclusion of grey literature and the dataset’s temporal limitation provide further challenges. In rapidly evolving domains where adaptation strategies continually adapt to new risks and technologies, the lag in data inclusion may render some insights outdated or incomplete. This temporal gap also limits the ability of GAMI outputs to inform real-time policy and practice.

These challenges mirror those noted by previous global reviews that have highlighted the difficulty of defining the scope and boundaries of large-scale adaptation evidence syntheses^5,33, systematic geographic, sectoral, and epistemic biases in climate change research and climate syntheses^34,35, and persistent inequities in who has the time, funding, and institutional support to participate in global collaborations^35,36,37.

To advance the next generation of GAMI, recent innovations in artificial intelligence (AI) offer promising new directions. AI can enhance the synthesis of peer-reviewed, grey, and media-based adaptation knowledge. Natural language processing is at the forefront of methodological innovation, enabling the extraction of high-level adaptation trends from large corpora of scientific and grey literature, while also enabling fine-grained, context-specific coding³⁸. For instance, recent machine learning approaches have mapped climate policy studies across diverse textual sources³⁹, others have assessed news articles⁴⁰ and the utility of language models in the coding process⁴¹. Moreover, AI can democratise access to climate data and modelling tools, particularly in Low- and Middle-Income Countries (LMICs)⁴², and offer opportunities to include evidence from diverse knowledge systems⁶. Emerging frameworks propose sustainability-oriented standards for AI applications, including transparency, reproducibility, and energy efficiency³⁸, ensuring that evidence is both machine-readable and ethically produced. Standardised, AI-compatible reporting protocols could enable dynamic, ‘living syntheses’ that also translate complex climate data into actionable insights for non-academic audiences, including policymakers and the general public.

Despite these advances, AI-assisted synthesis remains fundamentally dependent on high-quality ground truth data, underscoring the continued importance of expert human coders for training, validation, and interpretation. Large language models introduce epistemic risks, including hallucination, over-generalisation, and spurious pattern detection, which are particularly problematic in evidence synthesis contexts where traceability and factual accuracy are essential. Fully automated approaches therefore remain ill-suited for normative judgments and the interpretation of context-dependent or contested adaptation concepts⁴³. Hybrid human–AI workflows, combined with transparent documentation of models, training data, and validation procedures, are thus critical to ensure the credibility, reproducibility, and policy relevance of AI-enabled adaptation evidence synthesis.

Overall, GAMI has demonstrated the potential of community-driven efforts as well as the frontiers with regards to equitable knowledge production and synthesis. To create the enabling conditions for a more comprehensive understanding of the current knowledge on global climate change adaptation progress, a key priority towards a ‘GAMI 2.0’ is to establish a funded coordinating body that expands the adaptation evidence synthesis community, coordinates decentralised initiatives, provides systematic training for adaptation evidence synthesis researchers, supports researchers and non-academics from LMICs in the evidence synthesis process and promotes strategic knowledge transfer and communication activities. This includes co-design with researchers and practitioners from diverse and under-resourced contexts, equitable authorship practices, and mentorship for early-career scholars—principles aligned with recent calls for justice in global science⁴⁴, and particularly for regions that have received inequitably low investments in climate change research⁴⁵. While GAMI supported equitable authorship practices, it did not involve diverse perspectives at the conception and design stages, and future iterations should go further in their efforts to address the uneven burdens that time commitments place on global scholars. Advanced standards for reporting core dimensions of adaptation-related evidence immediately when publishing studies or grey literature on adaptation should be designed and communicated across communities of adaptation practices and research, to enable living evidence synthesis that can be consistently updated. These innovations point toward a future with accelerated pace of adaptation knowledge synthesis and enhanced inclusivity, accessibility, and policy relevance.

Methods

Survey method

We collected feedback through an online survey in May 2024 to gather insights from researchers who have experience with the GAMI process, the data, and further analysis of the data. The overall 129 researchers invited to the survey comprised all members of the GAMI coordinating, advisory team, and coding teams, and in addition, any lead authors of papers for which the original GAMI dataset was used, if they were not part of any of the GAMI teams. We used the online survey platform SoSci Survey⁴⁶ for a pilot run of the draft questionnaire, including six respondents, and to distribute the final questionnaire comprising eight open and 18 closed questions (with the option to elaborate responses in free text fields).

We analysed the collected data with the statistics software SPSS, using descriptive statistics to synthesise the responses around the four thematic blocks of the survey (see Supplementary Table 1 for the complete questionnaire): Respondents’ background; Evaluation of GAMI; Experience with GAMI spin-off papers; and Ways forward in adaptation evidence synthesis. In this way the survey results convey the experience with GAMI in three dimensions: (1) the direct experience of researchers in coordinating, conducting, and synthesising the review; (2) the perceived reception of the GAMI results in research, policy, and media arenas; and (3) the use of the GAMI data for further analysis and publishing, including additional peer-review feedback.

GAMI context

Our study builds on the personal experiences of the core researchers involved in GAMI. The GAMI team involves 129 people, of which eight comprise the coordinating team, 24 the advisory team, five the screening team, 97 the coding team, 20 the synthesis team, and three the expert solicitation team (people could have multiple roles). The process of GAMI to provide a global stocktake of human adaptation-related responses to climate change included searching literature in the scientific databases Web of Science, Scopus, and PubMed, as well as Google Scholar. Of the almost 50 thousand scientific documents on adaptation published in peer-reviewed literature between 2013 and 2020 screened, 1682 publications relevant to human adaptation responses to climate change were then coded according to various adaptation-related topics of interest (e.g., hazard, actors, response type, depth, speed, and scope) and synthesised around regions and sectors, amongst others (for a full description of the GAMI methods, see the published protocols^29,30,31). Besides the overall analysis of the GAMI dataset², it also provided the basis for further in-depth reviews and analyses on specific sub-topics, resulting in so far 18 published GAMI spin-off publications on diverse topics such as adaptation in mountains, islands, coastal cities, forests, Africa, in response to compound risks and extreme heat. Six out of 17 spin-off lead authors are from LMICs (35%).

Respondents’ background

The respondents—representing the membership of the GAMI team—are very mixed regarding career levels and disciplines. Most respondents are tenured faculty or at other career levels within academia. Eight respondents work outside academia, for example, as consultants or scientists in non-academic or governmental research institutions. Most respondents have a social science background (Fig. 7). Moreover, there is a strong dominance of respondents from and based in North America and Europe, very similar to the full GAMI author team (Fig. 8).

Fig. 7: Respondents’ academic background.

Survey respondents’ characteristics. a Career stage and b disciplinary background.

Fig. 8: Respondents’ geographic background.

Survey respondents’ country of primary institutional affiliation and origin compared to all GAMI authors.