Alan Liu

1. Overview
2. The Prompt
3. The Output
   1. Humanities Definitions Spreadsheet (on separate page)
   2. Description & Summary of Major Findings
4. Conclusion
   1. AI’s Conclusion (written by Fellou)
   2. Human’s Conclusion (written by Alan Liu)

Overview

This in-progress “Humanities Definitions Research Project” using the Fellou.ai agentic browser (home page | “quick tour”) was conducted as an experiment with agentic AI. Fellou is an early leader in the field of “agentic AIs” for research tasks (reviews of Fellou: example 1 | example 2). My ultimate goal with this project is to assist the new Center for Humanities Communication, which I cofounded, in gathering definitions of, and statements about, the humanities that can facilitate the development of training workshops and resources for communicating what the humanities are and do. But my immediate goal is just to test how useful agentic AI can be at its present level of development for research tasks that are difficult for humans to accomplish efficiently, systematically, or in well-planned ways.

Specifically, I asked Fellou to canvass a wide range of resources to gather definitions and statements about the nature, goals, values, fields, and scope of the humanities (see prompt below). This is only a prototype test in which I asked for 100 definitions or equivalent statements to be collected. Given the results I have inspected so far and their limitations (see my conclusion below), I think that collecting at least 1,000 items would be needed to ensure a wide representation of sources. But even at its current scale this is a good test of agentic AI because the blurred overlap between the academic and public domains of the humanities means that the canvassing I requested requires searching not just in well-bounded scholarly areas but also across a variety of popular, cultural, institutional, governmental, media, and other public or general audience domains that are either not well-bounded or not clearly relevant until actual instances are looked at. Even trained researchers would be hit-or-miss and grab-and-go in conducting such research. If I were to do this research solo (without either AI or a collaborative team or paid team of research assistants), I would likely only be able to sample areas and kinds of sources opportunistically, grabbing results that seem “interesting” or “important” based on what data scientists would call my “priors” or assumptions. The test here is to see whether, and how, Fellou can improve on human grab-and-go searching by widening and structuring search-and-collection strategies.

Why use agentic AI? The talent of such agents lies not just in planning and sequencing research actions in the general manner of generative AIs and their chatbots but also in autonomy of action. For example, while searching, retrieving, analyzing, excerpting, and categorizing materials that can include websites, PDF articles, and many others, Fellou can click through captchas and other interactive page elements (prompting the user for necessary account logins), fill out forms, and–perhaps most important–harvest input that it uses to initiate or redirect further chains of action. For example, Fellou can decide based on the amount or nature of information it finds that more research is needed in particular directions; and it can then plan and start those follow-on searches. In addition, the fact that Fellou is a downloadable program that runs on a user’s local computer as a browser (a “browser agent” working alongside a “file agent” and other agents) means that it can directly use local folders and files (see “What ‘computer use’ can do?”).

Technically, the Fellou program that a user downloads and installs is one instantiation of Fellou.ai’s Eko architecture–a JavaScript framework that allows developers to create cross-platform agent workflows concatenating the browser and file agents already mentioned with other ready-to-go or custom agents, and that also allows workflows to access for their AI work various large language and other generative models via OpenAI,  Anthropic, and other APIs. (See Eko Docs home page | “architecture overview”). In its version as a locally running browser program, Fellou accepts chatbot-like prompts to plan action sequences, which it asks users to inspect before running. Because it accesses AI’s via API, as mentioned, Fellou is LLM-agnostic and can tap multiple large language models (e.g., the ones driving ChatGPT or Claude), though it does not transparently indicate which AI it is using at any time. While Fellou is running, it opens search-progress and summary report views (see screenshots above). (At the current time, the Fellou interface feels like a work in progress. In my experience, it can be non-intuitive and confusing.) Fellou offers a free plan with a quota of monthly “credits” (“sparks”) that can be used to run workflows. There are also “plus,” “pro,” and “ultra” subscription plans with higher credit quotas and other features (see pricing).

The Prompt

The prompt I gave Fellou through its chatbot function was as follows (represented in a streamlined form that does not include several iterations I made to add stipulations or correct problems):

Prompt given to Fellou.

Search the web, including scholarly resources such as those found in Google Scholar, Web of Science, JSTOR, and so on to collect 100 definitions of the humanities, including not only statements designated as definitions but statements about the essential mission, goal, values, and scope of the humanities. Include as the humanities both academic fields and a range of literary, historical, creative, and other activities and materials related to the broader, public experience of the humanities. Also include definitions and statements about the humanities that bring into the same discussion near-humanities areas and activities (such as language or culture oriented aspects of legal studies, medical studies, etc.) As the output of your search, create a spreadsheet in which you collect quoted excerpts of definitions and statements about the humanities. Include citations of sources, where citations include where possible author(s), titles of works, the publication or website venue, the date of publication when that is declared, and URL, DOI, ISBN, ISSN, or other identifiers. Also tag definitions and statements by nation (US, UK, etc.) whenever it is clear that they are specifically about the humanities in a particular national context.

Fellou’s Step-by-Step Process (as reported by Fellou):

Thinking: The user needs me to conduct extensive research to collect approximately 100 additional definitions and statements about the humanities from scholarly and journalistic publications. This requires searching through academic databases like Google Scholar, JSTOR, Web of Science, and similar sources. I need to extract definitions about the mission, goals, values, and scope of humanities, including both academic fields and broader public meanings. The research should cover literary, historical, creative activities, and near-humanities areas like culture-oriented aspects of law and medicine. Each entry needs full citation details and should be tagged by nation when context is clear. The results should be presented in a structured document or spreadsheet format. This is a comprehensive research task that requires deep search capabilities and structured data organization.

File Agent: Create structured document with research findings.

The Output

» Humanities Definition Spreadsheet:

» Description & Summary of Major Findings:

Fellou also created an executive summary of results with summaries and analyses of “major findings”:

Major Findings (summarized by Fellou).

Research Methodology

Sources Consulted:

• Google Scholar: Academic articles and citations
• JSTOR: Scholarly journal articles
• Project MUSE: Humanities publications
• Academic.edu: Research papers and preprints
• SpringerLink: Academic publications
• Institutional websites: Universities and cultural organizations
• Government sources: National endowments and policy documents
• Professional journals: Specialized humanities publications

Geographic Coverage:

• United States (42.9% of entries)
• International/Global perspectives (28.6%)
• Netherlands (12.2%)
• United Kingdom (8.2%)
• Germany, Croatia, India, Sweden (remaining entries)

Key Findings

Core Definition Themes

1. Human Understanding and Experience (Most Prominent Theme)

The humanities are consistently defined as disciplines focused on understanding human nature, culture, and experience:

“The humanities are academic disciplines that study aspects of human society and culture, including certain fundamental questions asked by humans.” (Wikipedia)

“All the studies that the British Academy champions and promotes are about humanity itself: humanity past and present alike, together with whatever thoughts, concerns and hopes about human futures that those studies provoke.” (The British Academy, UK)

2. Interpretive and Critical Methods

Emphasis on interpretative approaches rather than empirical methods:

“The humanities are frequently defined as fields of study outside of natural sciences, social sciences, formal sciences (like mathematics), and applied sciences. They employ methods that are primarily critical, speculative, or interpretative, often with a significant historical element.” (Wikipedia)

“The extraordinary prominence given to subjective actions such as judgments, intentions, understandings, evaluations, and interpretations. In the humanities, but not in other disciplines, mental acts such as these are both the object and the end of scholarship.” (Geoffrey Galt Harpham, American Literary History)

3. Cultural Preservation and Meaning-Making

Role in preserving and transmitting cultural values:

“The humanities play a supremely important role in preserving fundamental ideals and values, ethical concepts, and religious and cultural concepts concerning the interrelationship between self and the world.” (Albrecht Classen, USA)

“The humanities study the meaning-making practices of human culture, past and present, focusing on interpretation and critical evaluation, primarily in terms of the individual response and with an ineliminable element of subjectivity.” (Humanities Hub, USA)

Academic Fields Within the Humanities

Based on the collected definitions, the humanities encompass:

Core Disciplines:

• Philosophy
• History
• Literature and Language Arts
• Classics (Ancient Greek and Latin)
• Religious Studies
• Art History

Creative Fields:

• Visual Arts (painting, sculpture, photography)
• Performing Arts (theater, music, dance)
• Creative Writing and Poetry
• Film Studies

Interdisciplinary Areas:

• Cultural Studies
• Gender and Women’s Studies
• African-American Studies
• American Studies
• Comparative Literature

Near-Humanities Areas

The research identified several fields that incorporate humanistic approaches:

Medical Humanities:

“Medical humanities use methods, concepts, and content from one or more of the humanities disciplines to investigate illness, pain, disability, suffering, healing, therapeutic relationships, and other aspects of medicine and health care practice.” (Johanna Shapiro et al., Academic Medicine)

Legal Humanities:

“Legal systems derive legitimacy not merely from formal authority but from their alignment with deeply held moral convictions rooted in humanistic traditions.” (Dawinder Singh, India)

Digital Humanities:

“Digital humanities is an interdisciplinary field that combines computational methods and technologies with traditional humanities research.” (Academic Source)

Educational Mission and Values

Civic and Democratic Purpose:

“Because democracy demands wisdom, the National Endowment for the Humanities serves and strengthens our Republic by promoting excellence in the humanities and conveying the lessons of history to all Americans.” (National Endowment for the Humanities, USA)

Personal Development:

“The ultimate function of the humanities is to liberate the mind from its subservience to brute fact… the cultivation of individual freedom, beginning with the freedom to think for oneself.” (Geoffrey Galt Harpham)

Critical Thinking Skills:

“Humanities majors are sought after for their critical thinking and problem-solving skills. They are particularly adept at ‘soft skills’ such as written and oral communication, creative problem-solving, teamwork, decision-making, self-management, and critical analysis.” (Wikipedia)

Historical Evolution

Ancient Origins:

“Ancient Greece formed the basis of a broad education for citizens.” (Wikipedia)

Medieval Development:

“Roman Times developed into the seven liberal arts: grammar, rhetoric, logic (the trivium), and arithmetic, geometry, astronomy, music (the quadrivium).” (Wikipedia)

Renaissance Transformation:

“Renaissance Humanism (15th Century): Shifted the humanities to subjects to study, focusing on literature and history (studia humaniora).” (Wikipedia)

Modern Challenges:

“20th Century (Postmodernism): Challenged traditional views, seeking to redefine humanities in more egalitarian terms for a democratic society.” (Wikipedia)

Contemporary Challenges and Opportunities

Declining Enrollment:

“While the total number of Americans with humanities degrees doubled between 1950 and 2010, the proportion of degrees awarded in the humanities has declined. At Harvard, humanities majors dropped from 36% in 1954 to 20% in 2012; nationally, they fell from 15% in 1993 to less than 9% in 2022.” (Statistical data from Wikipedia)

Digital Integration:

“The digital humanities can serve as a laboratory and means for thinking about the state and future of the humanities at large.” (Patrik Svensson, Sweden)

Interdisciplinary Growth:

“The assumption that science and humanities are opposed is a ‘longstanding presumption’ that benefits neither side, suggesting they are ‘points on a web rather than discrete disciplines.'” (David Clifford, UK)

International Perspectives

United States:

Strong emphasis on democratic values, liberal education, and practical applications. Focus on civic responsibility and individual development.

United Kingdom:

Emphasis on cultural heritage, institutional tradition, and broad intellectual inquiry about humanity.

Netherlands:

Leading research on the history and methodology of humanities, with emphasis on pattern-seeking and comparative approaches.

Germany:

Philosophical foundations emphasizing hermeneutic understanding and cultural interpretation.

Other International Contexts:

Recognition of diverse cultural traditions and emphasis on cross-cultural understanding.

Data and Research Methodologies in Humanities

Nature of Humanities Data:

“Research data in the humanities are the most diverse of all scientific disciplines, encompassing almost any data on human activity, including newspapers, photographs, diaries, church records, court files, etc.” (Ljiljana Poljak Bili? and Kristina Posavec, Croatia)

Methodological Characteristics:

“Key characteristics of humanities data usage include: 1. Research objects are collected, not created by the researcher. 2. Analogue data are critical. 3. License restrictions are often imposed by data owners. 4. Context is of great importance.” (Academic research from Croatia)

Philosophical Foundations

Truth and Interpretation:

“The humanities prioritize understanding meaning, purpose, and goals, distinguishing themselves from natural sciences’ focus on causality. Knowledge in the humanities is often seen as a ceaseless process of inventing and reinventing context.” (Wikipedia)

Subjective Understanding:

“Understanding in the humanities involves sympathetic apprehension, a fusion of horizons, subjective intuition, and intellectual sympathy.” (Geoffrey Galt Harpham)

Pattern Recognition:

“A central theme of the humanities across civilizations and ages is the identification of patterns in texts, art, music, languages, literature, and the past.” (Rens Bod, Netherlands)

Conclusion to Fellou’s Report on Major Findings

This comprehensive research reveals that the humanities maintain a consistent core identity centered on human understanding, cultural interpretation, and critical analysis, while continuously adapting to contemporary challenges. The definitions demonstrate remarkable consistency across cultures and time periods in emphasizing the humanities’ role in:

1. Understanding human experience across time and culture
2. Developing critical thinking and interpretive skills
3. Preserving and transmitting cultural values and knowledge
4. Fostering civic responsibility and democratic participation
5. Promoting cross-cultural understanding and global citizenship
6. Integrating with other fields to address complex human challenges

The research also reveals ongoing challenges including declining enrollment, the need for digital integration, and questions about practical relevance, alongside opportunities for interdisciplinary collaboration and global engagement. The near-humanities applications in medicine, law, and digital technology demonstrate the continued relevance and adaptability of humanistic approaches to contemporary problems.

This report is based on systematic analysis of 100 definitions and statements collected from scholarly sources across multiple international contexts, with complete citations available in the accompanying spreadsheet.

The results from this initial experiment indicate that while the research performed by Fellou.ai is not as “comprehensive” as it typically promises (in output file titles or headers such as “Comprehensive Humanities Research Project”), it significantly aids the human researcher in widening and structuring search strategies in ways that can later be extended manually or with the further aid of AI. The structuring of searches and outputs (e.g., into types of definitions or statements about the humanities and key ideas about the humanities) is particularly impressive. Also impressive is Fellou’s reporting to the user about planned steps, its progress during a run, and the content and structure of its results. Such reporting allows the user to understand (approximately) what is happening as Fellou rapidly shows multiple tabs and thumbnails during its traversal of research sources.

That said, however, there are in my view several major potential problems for the user in using agentic AI as represented by Fellou. The principal ones I encountered are the following. (Disclaimer: as a new user of Fellou.ai, my understanding and experience of the Fellou agents and interface are still evolving, and I may not be aware of methods of using the tool or of configuring its features that provide effective workarounds to the below.)

• First, Fellou for this project indiscriminately paraphrased and quoted sources in its output without reliably indicating the distinction between the two modes. After inspecting the definitions of, and statements about, the humanities that Fellou collected, I realized that the AI was always accurate in the essence of what it reports. But a user cannot tell if a statement is a paraphrase that closely approximates the words of a source or is an exact quote from the source. This remains true in my experiments with this project even after I explicitly varied my prompt to Fellou to ask for clear distinctions between paraphrases and quotations. In effect, this is like the “hallucination” effect of generative AI in general; though it is more accurate in this context to call it a “paraphrase” effect because (as far as I am able to tell) the generative AI is fabricating not untrue statements but non-literal (i.e., not word for word) statements.
  ___However, I should note that I have had better success in other projects I have started with Fellou where the results appear to be exact quotations as requested. It may be that careful constraints are needed in a prompt to ask for exact quotations and nothing else. For example, instead of asking in this project for both definitions and statements about the mission, values, and so on of the humanities–and for Fellou to distinguish between them–I could break the project into separate ones that collect exact quotes and essential ideas. In other words, it may be that the way to use a very ambitious, wide-scope, and flexible tool like agentic AI is to counterbalance its ambition with prompts that are purposely modular and narrow in design. The paradox, though, is that such a strategy runs against the grain of why one would want to use agentic AI in the first place.
• Secondly, Fellou has a hard time discerning and reporting on nested relations between sources. For example, it will report that a particular university’s humanities center is the author of a definition of the humanities that begins, “The term ‘humanities’ includes, but is not limited to, the study of the following: language, both modern and classical; linguistics; literature; history;… [etc.].” But it will not note that the humanities center is actually quoting the legislative act that the National Endowment for the Humanities uses to define the humanities. For experienced human researchers, it normally goes without saying that when one source quotes another source there are actually two sources involved. Which one is assigned the role of “author” is a matter of context. For agentic AI, and for computational processes generally, such distinctions cannot go without saying. And even saying them in a prompt does not guarantee that the AI will listen.
• Thirdly, Fellou can concentrate on particular sources to mine multiple materials in bursts. For example, Fellou collected 14 separate paraphrases or quotes from Geoffrey Galt Harpham’s essay “Finding Ourselves: The Humanities as a Discipline.” This may or may not be a “feature,” rather than “bug,” depending on the needs of a researcher. But in any case it means that a  researcher needs to take care to ask Fellou to scale its searching well beyond the quota of 100 found items that I asked for in my prototype test to ensure that the results are representative of breadth. Fellou could usefully give the human researcher a better sense of how it chooses to collect from sources. For instance, is it looking for relatively short, discrete passages that pass the threshold of a relevancy score?
• Fourthly, the human user would develop a more effective cognitive map of the research territory traversed by Fellou if the reporting it generates also provided a sense of areas adjacent to those it looks into, thus allowing the user to see the overall research map being navigated by the AI and what parts of the map are, and are not, being investigated. (A visual flowchart or actual map-like visualization of research areas that the AI passes through, and also by, would be particularly intuitive in this regard.) This is a higher level kind of cognitive problem in AI-use (related to interpretability and explainability machine-learning problems generally) that is of course not specific to Fellou.
• Fifthly, human users would benefit from Fellou being more proactively transparent about the security implications of using agentic AI. Besides any possible technical issues of data security (see Anatoliy Fedorenko, “Security Analysis of the Fellou AI Browser,” Medium, July 24, 2025), the kinds of concerns I indicate have to do with the fact that Fellou has the power to traverse both online sites and local apps and files in ways that users would feel more comfortable about if they could in advance choose to “fence off” particular external or local resources. While Fellou requires the user to take action on a case-by-case basis to log into certain external sites, such piecemeal permissions do not make it possible to strategize research in advance around security concerns; and it also does not accommodate other kinds of “no go” user decisions that are not strictly related to security–e.g., a user’s wish to avoid a domain, site, or service for any practical, social, political, locality-based, or other reason.
• Finally, in its present stage of development, complex jobs with many steps can (in my experience) result in job-stopping errors that are undocumented, leading to two kinds of frustrations. One is that the user is left unclear about what has and has not been completed in the present step, and where transient or final deliverables have been deposited. The other is that the credits (“sparks”) used for a failed job are not returned to the user, and pressing “run again” typically starts a complex job over again from the beginning with the result that another large quota of credits is burned up.

In sum, as exemplified by Fellou, agentic AI has the potential to automate research by humans to increase the scale, speed, and structure of research activities that include searching, collecting, citing, categorizing, summarizing, and analyzing. By contrast with earlier kinds of large language models and AI chatbots (though these are quickly also evolving agentic capabilities), agentic AI can “see” and “act on” online and local digital resources in ways that are more multifarious, autonomous, and logically sequenced toward a goal. The fact that Fellou is LLM-agnostic and can use multiple AIs in the background is also potentially an advantage. But counterbalancing these advantages are limitations such as those described above that make it difficult for humans to align their understanding with the AI’s of what the machine agent is, or should be, doing.

In the case of my specific research project, the definitions of, and statements about, the humanities returned by Fellou are ultimately valuable despite, or sometimes because of the limitations I mention. The payoff is a set of short “definitions” and “statements” that are useful for two reasons in particular–at least in my estimation as a long-time scholar working on advocating and communicating the humanities (e.g., through 4Humanities.org, the WhatEvery1Says project, and the Center for Humanities Communication).

One reason is that the paraphrases/quotes Fellou gathered have a cut-to-the-chase succinctness. The brevity of the excerpts collected in column B of the Humanities Definitions Spreadsheet created by Fellou is an antidote to the usual wordiness of the humanities. As we found in the WhatEvery1Says project, which condensed its key findings to 1-page “cards” (examples), it is enormously clarifying to complement complex thoughts about the humanities (usually expressed by humanities scholars in articles or books) with much briefer statements focused on essential points. The outcome of this Fellou experiment is a set of pithy, condensed ideas about the humanities.

The second reason is that the “definitions” collected by Fellou are non-uniform in their structure, format, voice, and grammatical forms in ways that may at first glance seem the opposite of what is desired. But this heterogeneity is actually in my view a virtue. My experience over many years of researching and thinking about the idea of the humanities has taught me that looking for cut-and-dry definitions of the humanities is a fool’s errand. Some definitions are declared as such, and even itemized point-by-point. But many other de facto definitions take the form of a whole variety of statements–in different forms and voices–about the mission, goal, values, and scope of the humanities. Agentic AI is well-suited to looking for such heterogeneity.

Considering in overview the definitions/statements about the humanities gathered by Fellou in this experiment, I see value not in the form of a final product but in what amounts to raw material for follow-on products. I feel that I have made a start on gathering and distilling ideas, keywords, phrases, and mottos about what the humanities are and do that can be a resource pool for evolving further, possibly more publicly impactful visions of the humanities in the future.

For example, after completing my agentic AI test I asked another, generative AI (Anthropic’s Claude Opus 4.1 in this case) to ingest the Humanities Definitions Spreadsheet produced by Fellou and look at its columns B, N, and P (“definitions,” “values emphasized,” and “keywords”) to create five different “vision statements” about the humanities. The following is the result, which I would not use “as is” but instead as guidance on how I and others might write these even better.

Prompt given to Claude Opus 4.1:
The attached spreadsheet is the output of a research project that collected approximately 100 definitions or statements about the nature, mission, values, and other features of the humanities. Analyze the spreadsheet, paying attention especially to columns B, N, and P. Then write five different vision statements on behalf of the humanities. The vision statements should be relatively short (a few sentences or up to a paragraph), and they may (but need not) be in a format that includes bullet points.

Claude’s response:

I’ll analyze the attached spreadsheet to understand the various definitions and perspectives on the humanities, then create five distinct vision statements based on the patterns and themes I find. Based on my analysis of the spreadsheet’s approximately 100 definitions of the humanities, I can see several recurring themes: understanding human experience and meaning, preserving cultural heritage, fostering critical thinking and wisdom, promoting creativity and imagination, bridging disciplines, and addressing societal challenges. The values emphasized most frequently include understanding, knowledge, creativity, justice, and engagement, while keywords consistently point to history, philosophy, literature, culture, and human experience.

Here are five distinct vision statements for the humanities:

Vision Statement 1: The Bridge Between Past and Future

The humanities serve as humanity’s living memory and imagination, preserving the wisdom of our ancestors while cultivating the critical thinking needed for our collective future. Through the study of history, literature, philosophy, and the arts, we maintain an unbroken conversation across generations, cultures, and continents. We are the stewards of human meaning-making—interpreting where we’ve been, understanding who we are, and imagining who we might become.

Vision Statement 2: The Laboratory of Human Understanding

The humanities constitute a unique field of inquiry dedicated to exploring the fundamental questions of human existence: Who are we? How should we live? What does it mean to be human? Our mission is threefold: • To investigate and interpret the vast diversity of human experience across all cultures and throughout history • To develop wisdom that transcends mere information, fostering the capacity for critical reflection and ethical reasoning • To illuminate connections between individual lives and broader social, cultural, and historical contexts

Vision Statement 3: The Creative Force for Social Transformation

The humanities stand at the intersection of imagination and justice, creativity and community. We believe that understanding human culture, values, and expression is essential to building more equitable and meaningful societies. By engaging with literature, philosophy, history, and the arts, we cultivate empathy, challenge assumptions, and envision alternative futures. The humanities don’t just describe the world—they help us reimagine and reshape it.

Vision Statement 4: The Interdisciplinary Catalyst

In an era of complex global challenges, the humanities provide essential perspectives that complement and enrich scientific and technological approaches. We are: • Pattern-seekers who trace meaning across texts, artifacts, and human creations • Bridge-builders who connect disparate fields of knowledge through shared human concerns • Translators who make specialized knowledge accessible and relevant to diverse audiences • Collaborators who bring historical context, ethical reflection, and cultural sensitivity to grand societal challenges

Vision Statement 5: The Cultivation of Complete Human Beings

The humanities are dedicated to nurturing not just knowledge but wisdom, not just skills but understanding, not just careers but meaningful lives. Through engagement with philosophy, literature, history, languages, religion, and the arts, we develop the whole person—fostering creativity, cultivating judgment, and deepening our capacity for reflection. We prepare individuals not merely to navigate the world as it is, but to question, interpret, and ultimately transform it through the power of human imagination and understanding.