third-submission.tex

\documentclass[10pt,letterpaper]{article}
\include{settings}

\newcommand{\rulemajor}[1]{\section*{#1}}

\begin{document}
\vspace*{0.2in}

\begin{flushleft}
{\Large
\textbf\newline{Ten Quick Tips for Making Things Findable}
}
\newline
\\
{Sarah~Lin}\textsuperscript{1,*},
{Ibraheem~Ali}\textsuperscript{2}
{Greg~Wilson}\textsuperscript{1}
\\
\textbf{1} RStudio, PBC
\textbf{2} Louise M. Darling Biomedical Library, University of California, Los Angeles
\\
\bigskip
* sarah.lin@rstudio.com
\end{flushleft}

\section*{Abstract}

The distribution of scholarly content today happens in the context of an immense
deluge of information found on the internet. As a result, researchers face 
serious challenges when archiving and finding information that relates to their work.
Library science principles provide a framework for navigating information ecosystems
in order to help researchers improve findability of their professional output. Here
we describe the information ecosystem which consists of users, context, and 
content, all three of which must be addressed to make information findable and usable. 
We provide a set of tips that can help researchers evaluate who their users are, 
how to archive their research outputs to encourage findability, and how to leverage 
structural elements of software to make it easier to find information within and beyond
their publications. As scholars evaluate their research communication strategies, 
they can use these steps to improve how their research is discovered and reused.

\section*{Author summary}

Sarah Lin is the Information Architect and Digital Librarian at RStudio, PBC.
Ibraheem Ali is the Sciences Data Librarian at the University of California, Los Angeles.
Greg Wilson works in the Education team at RStudio, PBC.

\section*{Introduction}

Researchers have always had to manage information, but the exponential growth of
electronic data has both required and fostered the creation of new ways to do
this \cite{Rosenfeld2015,Hedden2016}. The problem is not just about finding a
particular article or website, but also how to find the information within that is 
important for reusability and interpretation of the work. Information may be stored
in many formats, exist in multiple versions, and need to be shared with varied audiences
for both research and teaching.

In the last 4 years, growing adoption of FAIR principles \cite{Wilkinson2016}
has helped researchers manage the data and other digital objects associated with
their professional work. The FAIR principles encourage data to be Findable,
Accessible, Interoperable, and Reusable. Indeed, research reproducibility
hinges on these four principles, which form the backbone the following 10 tips.
This paper is concerned with findability, including research data and
extending beyond it to encompass all types of work products produced during a
researcher's career, inside or outside academic settings.

Library science offers ways to work through the maze of information generated in
professional life, and librarians' skills can be applied by any researcher who
seeks to improve access to -and utilization of- their research and other 
professional outputs. The ten quick tips in this paper build on the fact that all
information ecosystems have users, context, and content \cite{Rosenfeld2015}.
To solve the information retrieval problem, researchers must therefore think
broadly about who needs that information and the context within which it is
created, as well as its actual content.

\rulemajor{1. Design for a wide range of users.}

The first step in making information findable is to determine who will be doing
the finding. This includes everyone who might learn from your work, contribute
to it, expand upon it, or re-share information through their own networks
\cite{Covert2014}. While you might think there are only a few relevant experts
who know your field well, novices and trainees will also need to use your work
as they gain experience in the field, thereby making your actual user base
considerably larger and more diverse. Furthermore, some users need to access
scholarship through an intermediary, such as translation software or a screen
reader. Being mindful of all potential users and how they might need to
interact with you and your work is the foundation of all ten tips.

The information you wish to convey and the way it is currently organized may
make perfect sense to you, but its meaning for your users is determined by what
\emph{they} interpret from the information they encounter and the way it's
arranged. This means that the organizational strategies you use are a
communication channel in their own right. To illustrate this, Borges created a
classification of animals whose categories included ``those belonging to the
Emperor,'' ``embalmed ones,'' ``suckling pigs,'' ``those included in this
classification,'' ``those drawn with a very fine camel hair brush,'' and ``those
that look like flies from far away'' \cite{Borges2000}. While this was
deliberately ridiculous, it illustrates the fact that every way of organizing
knowledge embodies choices by the organizer, which may or may not align with
those of the audience. Therefore when preparing materials for sharing it is
important to establish context, to use clear and concise language, and minimize
the use of jargon.

More concretely, consider the website of a faculty member coming up for tenure:
she created the website as post-doc to publicize her papers and to make it
easier to fill out grant applications by listing professional activities in one
place. Yet the site might be useful in other cases as well:

\begin{itemize}

\item
  Colleagues may come to the site looking for un-paywalled copies of her papers,
  to find out what she's currently working on, or where she is next going to
  present her work.

\item
  Tenure committee members might review her accomplishments to assess her
  work's impact.

\item
  A librarian (or a program written by a librarian) might scrape that site for
  journal articles to include in the university's institutional repository.

\item
  A student might come to the site looking for course information or materials.

\end{itemize}

Reaching out to a variety of users with distinct needs to ask for findability
feedback will help you discover any gaps in organizational alignment.

\rulemajor{2. Design with the end in mind.}

Given the current state of technology it can be easy to rapidly create digital
information. With the plethora of software and formats you may employ, you
almost certainly have information in lots of different formats, file types and
locations. The second step in making things findable is to think ahead about
the things you would want to be found at the completion of the project before
cataloging what you have. That way you can have time to anticipate what can go
where, and adapt if necessary before materials are published.

This is particularly relevant when examining a typical research workflow. A
researcher uses a particular set of data sources (Figure~\ref{workflow}A), such
as model organisms, molecular systems, publicly available next-generation
sequencing data, or other locally curated collections of information.
Subsequently, data sources are treated with specialized protocols and tools in
the lab (Figure~\ref{workflow}B) which help visualize, identify or extrapolate
new observations about biological phenomena (Figure~\ref{workflow}C). After
repetition, trends seen among noisy biological observations can be further
analyzed, visualized, and statistically evaluated with software or code
(Figure~\ref{workflow}D). Finally, the researcher builds context for the work by
writing a manuscript and citing relevant literature (Figure~\ref{workflow}E).
Manuscripts are revised by colleagues in the field through peer review
and published in a peer reviewed journal.

\begin{figure}
  \includegraphics[width=\textwidth]{workflow.png}
  \captionsetup{justification=centering}
  \caption{A typical research workflow. A-E: Major components of the research workflow.
  F: An example set of citable repositories or tools. G: Persistent identifiers used
  for the example repositories listed. DOI: Digital Object Identifier, RRID: Research
  Resource Identifier, ROR: Research Organization Registry, ORCID: Open Researcher and 
  Contributor ID. * Some institutions maintain their own repositories for archiving data
  and may be available to researchers at a low or no cost}
  \label{workflow}
\end{figure}

By planning ahead, you can identify what research products can go where, and
update them if necessary before materials are published in a peer-reviewed
journal or cited in a grant application. Advancements in best-practices
for making research products more `FAIR' has led to the creation of an array
of discipline specific, and general repositories\cite{PLOS2020}. Researchers
can deposit research products created from each of the steps in the workflow (Figure~\ref{workflow}F).
Many repositories now create Digital Object Identifiers (DOIs)\cite{DOI2020}, or
other permanent identifiers for their submissions, making them citable and
easily linked to the contributing researchers via Open Researcher and Contributor
IDs (ORCIDs)\cite{ORCID2020} (Figure~\ref{workflow}G). Many repositories follow the standards
set by experts in the field, reporting information that is recommended for
reproducibility while employing necessary practices to protect the privacy of
sensitive information.

Elaborating on our example of the hypothetical faculty member. She wants to plan 
ahead to be sure her research is accessible and reusable by her lab, students and 
collaborators. She works in a competitive field, so she also wants to maintain 
some privacy with regard to works in progress. She takes advantage of some
repositories and tools that help make her work easier to find for a variety of
users (Figure~\ref{workflow}F). This way nearly all of her major research products can be citable and
are interlinked. The resulting network of citations aids findability greatly:

\begin{itemize}

\item
  Finalized laboratory data is stored in Dryad\cite{DRYAD2020} which allows her 
  to cite the same data source in multiple publications with a single DOI. She can even 
  update the data archive as new pertinent data is collected and the link will not change.
 
\item
  Lab protocols maintained in Protocols.io\cite{Teytelman2016} ensure future
  collaborators, lab members and students can easily find her established
  protocols using a DOI, while keeping incomplete protocol drafts private.

\item
  Code used to analyze the data is deposited in Zenodo\cite{ZENODO2020} which
  she can cite and update as code gets optimized and new versions are created,
  which also allows anyone looking to reproduce her research to also reproduce
  the analysis.

\item
  Posting finalized drafts of her student's research manuscripts on a preprint
  site like BioRxiv\cite{bioRxiv2020} enables colleagues to download the draft
  easily, discuss in an upcoming Journal Club, and publish the feedback publicly PREreview\cite{PREreview2020}.
  
  \item
  Her ORCID identifier is linked to each of the DOIs created by these
  repositories, ensuring that students, collaborators, tenure committees, and
  librarians can all access her research outputs with one link.

\end{itemize}

\noindent
Following the standards recommended by experts\cite{Wilkinson2016} in the field
as you design your information ecosystem will reduce the barrier to finding
relevant materials associated with your work. Furthermore, most publishers already
require data and code archiving to encourage reproducibility. Remember that your future self is
also one of your users: everyone is prone to forgetfulness, so anything you do
for others will likely pay off for yourself eventually\cite{Briney2015}.

Planning to incorporate persistent identifiers into the research workflow is a
straightforward way to design for findability. Remember too that the strategies
users employ to find something can look different in different contexts. Beyond
finding your published or unpublished work on the web, users may need to find a
specific item within a website, and/or find a particular piece of information
within a specific file or item within a webpage. Depending on your content, you
may have challenges in all three areas; the remaining tips on context and
content will help you address these issues.

\rulemajor{3. Use textual structure.}

Findability at the document, post, or article level can be improved by taking
advantage of the textual structures that information management programs provide
\cite{Hedden2016}. For example, a key part of searching the web is scanning the
text returned by search engines to see if it contains target information.
Textual structure helps that process \cite{Krug2014}: formatted headers (rather
than just enlarged text), bulleted or numbered lists, and \textbf{highlighting}
terms that are important all make both the information and its structure easier
to understand. Similarly, headings and table of contents can be hyperlinked,
which supports both scanning and navigation. Textual structure aids navigation
both by helping users create a mental map of the webpage or document they have
found, but also by exposing elements utilized by screen readers to make your
work accessible.

Textual content is created and aggregated in so many forms, using so many
different programs, that it is difficult to specify strategies beyond headings,
lists, and highlighting. However, specialists working in the same field tend to
adopt the same tools, so it is worth exploring how your peers annotate
information as well as creating, manipulating, or storing it. For example:

\begin{itemize}

\item
  GitHub allows users to add tags to issues and commit messages which can then
  be searched across projects.

\item
  Electronic lab notebooks can use XML schemas like Darwin Core, EML, or FITS
  \cite{Briney2015}.

\item
  Using specific Google Docs heading levels creates a table of contents in real
  time, visible when the file is open.

\item
  CSV files do not have a standard way to store metadata, but authors commonly
  created a README or MANIFEST file that describes the structure and content of
  the files in a collection. (See \cite{Pudding} for examples.)

\end{itemize}

On a practical level, templates for file creation, data collection, and electronic 
lab notebooks makes it easier to be consistent and to spot inconsistencies.

\rulemajor{4. Add metadata.}

Just like people who end up with piles of photographs with nothing written on
the back, we all have digital mounds of files and content with no metadata
describing when it was created or what it contains. Even the most basic
metadata provides extra clues for information retrieval; however, what you can
add depends on the software you use to create, store, and access your
information, and on the file formats that information is stored in.

Almost all modern operating systems allow you to add information to the
Properties of a file or directory. Databases, word processors, and website
construction programs also have built-in metadata capabilities, though they may
be hard to find and harder to understand how to leverage. To make matters
worse, the fact that metadata is often software-specific makes it easy for
inconsistencies to creep in. For example:

\begin{itemize}

\item
  The tags used on a WordPress website may not be in step with the properties in
  the images on that site.

\item
  Keywords added to a journal article when submitting to the publisher's site
  are not automatically added to the metadata in the PDF being submitted.

\item
  When a citation is copied from a database or repository to a bibliography
  manager, the software may not copy over the structural information implied by
  the article's location in the database.

\end{itemize}

The most difficult thing about metadata, however, is getting into the habit of
creating it in the first place. If you get to choose what software to use, it
helps to pick one that simplifies metadata creation. For example, most website
generators allow you to type tags into an article's header without having to
define them first. This can lead to a proliferation of synonymous (or
misspelled) tags, but some occasional cleanup is better than tackling a mountain
of untagged information. Repositories that force metadata creation upon
submission greatly assist efforts to make work product findable, and researchers
would be well served to replicate those metadata elements within their own file
storage schemas. Indeed, creating an internal taxonomy (list of terms) or
ontology (list of relationships) at the beginning of a project can make
assigning metadata much easier.

You should also examine how metadata can be transferred from an old system to a
new one if you have the luxury of switching software (or have a change forced on
you). Some form of XML is usually the best option when doing this: it is likely
to be with us for many years to come, and the same pedantry that makes it
tedious for human beings to type and read ensures that programs can read it
without having to guess what its creators actually intended. FAIR principles
help ease the burden of software and storage migration. They encourage
researchers to plan for interoperability in software and data storage from the
beginning of their projects, and reduces concerns about data migration.

\rulemajor{5. Use search \emph{and} browsing}

Research on information seeking shows that people search \emph{and} browse when
they're trying to find information. As they browse a website, document or
file, they build a mental map of the content they could possibly find, then
search based on that map. ``In the process, they modify their information
requests as they learn more about what they need and what information is
available from the system'' \cite{Rosenfeld2015}. You have probably seen or
done something similar with a print book, trying to determine if it's one you
want by looking at the table of contents and the back cover. These two
functions work together because search allows users to find information they
know they need, whereas browsing allows users to find information they don't
know that they need\cite{Bates2002}. Designing for both browsing and searching
is especially pertinent in a search algorithm environment that often dynamically
creates results unique to each individual search executed, influenced by the
user's previous interaction with a particular website and/or internet browser.

You should therefore make information accessible both ways and make it easy to
move from searching to browsing and back again. Tags and other metadata help
with searching, while structural clues tell users about the content contained in
the information they are looking at. That communication, ``enables the answers
to users' questions to rise to the surface and answer questions like, Where am
I? What's here? Where can I go from here?''  \cite{Rosenfeld2015}. Similarly,
``{\ldots}the words you use in the navigation systems and headings of [your
  content] help you find what you're looking \emph{for}, but they also help you
understand what you're looking \emph{at}'' \cite{Arango2018}.

For example, when users don't know exactly what they need, the terms in a menu
help them understand the vocabulary used in this domain and the boundaries of
what is included (i.e., terms are listed) or excluded (i.e., no menu terms
exist). At the same time, the headings in the documents they find act as
topical markers: they help users summarize the information contained in the
document, but also refine what they would search for based on the terms used in
those headings. Navigation bars on websites function in a similar way: if the
user knows exactly what they are looking for, they can scan the menu and select
the option that matches their need.

Additionally, users might want content organized chronologically or topically.
In websites that are generated programmatically using tools such as
Blogdown\cite{Xie2017} or Wordpress \cite{Williams2015} these types organization
features may be simple to implement. However, increasing the number of site
navigation options may confuse the user if it is unclear how items relate to
each other. Authors must strike a balance between offering filtering and
searching options, and being as inclusive as possible when deciding how
materials will be structured and shared with their users.

\rulemajor{6. Mimic real world directions.}

The language we use in digital environments mirrors that used for physical
directions: we ``visit'' or ``go to'' a website without actually changing our
physical location. Using the navigational metaphor consistently helps users
build the mental map mentioned in the previous tip. File paths and breadcrumb
trails on websites give users a sense of where the information resides and
suggest new paths they can take \cite{Krug2014}. For example, the URLs of a
website might all include the name of a section of the site, such as
\texttt{/papers/} or \texttt{/blog/}. DOIs and ORCIDs function as the best
kind of internet `signage,' ensuring users never encounter a``webpage not found"
error message in the course of retrieving a particular publication.

While much has been written about web usability in general
\cite{Covert2014,NNG2020}, library science focuses on information-seeking
behavior. For example, we know that users scan but don't read: they click on
the first close thing they see and give up very, very quickly \cite{Bates2002}.
Your markers and directions should therefore be as consistent as highway signs
with regards to appearance, style, and type of information. Wherever possible
(and it's \emph{always} possible), use mechanisms that users will have become
familiar with elsewhere, such as the vertically nested folders of file browsers
or the left-to-right arrangement of breadcrumb trails. Persistent identifiers
are a great example of markers that consistently aid navigation in structured
(e.g., databases) and unstructured (e.g., personal websites) environments.

\rulemajor{7. Use meaningful names.}

The names of files and URLs of webpages are the one piece of metadata you cannot
avoid creating, so always choose ones that are human-readable and that convey
information about what they name, both when navigated to \emph{and} when
returned in search results. Returning again to the faculty member's website, it
would be easy to name a paper \texttt{plos2020.pdf}, but since other people may
also have published papers in PLoS in 2020, a more structured name such as
\texttt{lin-findability-plos-2020.pdf} will both convey more information at a
glance and retain that information after the paper has been downloaded and put
in a folder with dozens of others.

There are many ways to develop a naming schema, largely related to the nature of
the information you create. At the most basic level, ``you should use
consistent names for the same reason that you use good file organization: so you
can easily find and use data later. Additionally, good naming helps you avoid
duplicating information \cite{Briney2015}. Researchers with multiple research
projects or significant complexity in their data sources should establish and
document a unified system of abbreviations for those projects or sources; these
can be summarized in a data dictionary or README file. Consistency is key:
standardizing on lower case, a preferred date format (YYYMMDD or YYYY-MM-DD will
both sort chronologically), and filename suffixes (\texttt{.jpg} instead of
\texttt{.jpeg}) will help everyone find what they need
\cite{Wilson2014,Wilson2017}.

Renaming existing files to be consistent with your standards after the fact can
seem like a waste of precious time, but since the research cycle doesn't end
with publication \cite{Briney2015}, there is a very high likelihood that someone
will need to reuse your data and will have to try to figure out what files
corresponded to what part of your research. Similarly to establishing metadata
norms before beginning a project, creating a naming convention that is adopted
by any collaborators before research begins will preserve findability into the
future.

If you have things to name that are not files, such as projects, web pages, or
document headings, remember that the more generic a term is, the harder it is to
search for: naming a raw data file ``raw'' or a downloaded file ``download''
makes finding the information they contain nearly impossible. A quick test is
to search for the name before adopting it: if dozens of unrelated results come
up, you may want to pick a different name. You should also think about
nicknames or shortened versions of your names and make sure they are present in
text or tags so that the content can be discovered by a search engine and a
user.

\rulemajor{8. Use tags.}

After meaningful names, tags are the easiest and most effective metadata you can
create. Almost all digital tools allow users to add arbitrary tags to items:
file properties on Windows and labels on GitHub issues are just two examples.
Additionally, almost all search tools leverage tags to narrow a query's scope.
This means that you can now file a single thing in multiple ``locations'', which
was not possible in the pre-digital era. Multiple tags also assist users from
varied backgrounds because the terms can be customized to be inclusive of a
diverse set of users.

When choosing tags, be consistent in your depth of topical term assignment (how
specific your terms are) and your selection of terms for subject and format (the
number of terms you use to describe each subject and format). For example, if
you tag some items in an ecological data set with a species name, don't tag
others simply as ``reptile'' unless the species is unknown, in which case you
should:

\begin{itemize}

\item
  tag all items ``reptile'', ``bird'', ``mammal'', and so on for high-level
  searches, and

\item
  tag all items with a species, which might be ``unknown'' or ``NA'' (not
  available), or
  
\item
  tag all items at both general and increasingly specific categories if that is
  the standard for your discipline \cite{FAIR2020}.

\end{itemize}

What should you tag?  The answer is ``everything'' from informal personal notes
to data sets submitted with publications or included in repositories, because it
is all material you will want to be able to find later. The benefit of tagging
comes from doing it in all of those situations, not just when a journal
submission requires it.

If you are certain something is for purely personal use, you can create your own
taxonomy of subject keywords, which is called a \emph{folksonomy}. Folksonomies
are what you see with tags on Flickr: early content creators assign terms as
they see fit, and later contributors can use those or add their own. If you
take this route, it's worth reviewing new tags regularly to look for synonyms,
misspellings, differences in capitalization, singular/plural discrepancies, and
other inconsistencies.

What terms you use as tags for personal consumption may not matter much, but
work that is shared with colleagues should use particular terms or tags that
conform to relevant standards \cite{FAIR2020}. These terms typically come from
taxonomies, thesauri, and ontologies: taxonomies and thesauri generally have
built-in subject hierarchies that can help you create navigational structure,
while ontologies map relationships between ideas. Crucially, all three are
\emph{controlled vocabularies}: they are a defined list of terms created and
maintained by experts rather than being crowdsourced like a folksonomy.

There may or may not be relationships built between terms in a controlled
vocabulary, such as equivalencies (``CA'' for ``California''), broader/narrower
terms (United States/California), and/or replacement (weed \emph{use}
marijuana). Established subject terms will match article databases, data
repositories, and library catalogs that you and your users might already be
familiar with, which will again aid search and navigation. Well-known examples
in the United States include the National Cancer Institute (NCI) Thesaurus
\cite{NCI2020} and the Medical Subject Headings (MeSH) \cite{ASI2020}.

\rulemajor{9. Understand the difference between format and subject.}

However you create tags, you need to address the distinction between format and
subject. Format describes what your content \emph{is}, while subject describes
what it is \emph{about} \cite{Joudrey2015}. About-ness is the most common
content analysis, but is-ness issues will probably affect people's ability to
use your information, so you may want to add metadata to make it explicit.

A simple example of this is a blog post on a researcher's professional website.
The post is \emph{about} a subject, like a book review, but it \emph{is} a blog
post rather than biographical details, bibliography, or a list of currently
taught classes. Going back to your users, what subjects are important to them?
And do those topics carry over or change between differences in format? For a
librarian, this is basically a question of combined terms: are your format terms
uniquely matched to topics (e.g., blog posts are always about news) or do you
have multiple topics in each format (e.g., blog posts and tutorials on the same
subject)?

Similarly, you can rely on filename suffixes to distinguish computational
notebooks from PDF files, tabular data sets, or slide decks, but should use
tagging, a filename convention, or a description in a README to tell people
whether the contents are raw information, tidied-up data, or an aggregation of
several underlying datasets. This enables users to search by topic, format, or
both.

Since dissemination sometimes changes a file's format (e.g., printing slides to
a PDF), naming and metadata conventions tend to be more robust as well as more
informative than relying on file types. Once again, structural clues can help:
a folder specifically for conference presentations may contain one sub-folder
for each presentation, which in turn contains the PowerPoint and PDF versions of
the presentation with exactly the same names but different filetype suffixes.
Likewise, journal articles you store will need a naming or structural convention
to distinguish articles you have written from those you have downloaded for your
own use.

\rulemajor{10. Do not abbrvt.}

Acronyms and abbreviations make communication between those who know them more
efficient at the price of making them less accessible to newcomers. Spelling
out acronyms and abbreviations that you take for granted (or hyperlinking to
their definitions) makes information easier to find and enables newcomers to
participate in conversations that are considered technical or advanced. When
doing this, remember that acronyms are often repurposed by different professions
or disciplines: what seems obvious to you is probably not obvious to people from
other communities. Since every discipline has some common abbreviations, write
them all out in full the first time they appear or create or point to a term
dictionary.

\section*{Conclusion}

Changing work habits is hard, so remember that while perfection isn't possible,
progress is. Start by deciding whether to begin your next project with a new
set of information organizing principles or to go back and alter existing
artifacts \cite{Briney2015}. You might also consider this process as you would
a research experiment, and incorporate one small change at a time. Whichever
you choose, the ``ways you enforce your way of doing things changes how users
think about the place[s] you made and perhaps ultimately, how they think about
you'' \cite{Covert2014}.

\bibliography{10-findable}

\end{document}