How to conduct a literature search

Objectives:

• Normalize struggle inherent in developing literature skills, especially around tasks the seem trivially easy but are in fact difficult (like googling the correct keywords)
• Promote a growth-mindset: this is a skill that is learned, not inborn.
• Teach students how to break-down the literature search into smaller, simpler task – see guide on how to conduct a literature search.

Pre-class work

• Read Et al. for all: Citations as a Tool for Racial Equity, Inclusion, and Justice, and answer:
  • What are the negative effects of citation bias?
  • What can we do to adopt more conscientious citation practices?
• What do you find to be the most challenging aspect of doing a literature search?
• Describe your current strategy for doing a literature search for each of the following components: finding papers, reading papers, and tracking papers.

In class [slides]

This unit generally follows the structure of our guide on how to conduct a literature search.

1. [15min] Students introduce themselves with the slides they made
2. [15min] Introduction: recap of last class, group discussion of reading about citation biases
   • What were everyone’s takeaways from last class about how to skim a paper?
   • Has anyone tried this strategy over the last week? How did it go?
   • What are some takeaways from the pre-class work about citation biases?

     Insight: (a) Citations are often used as a measure of success, problematically leading to inequities. (b) More broadly, everything we do as researchers leaves a fingerprint – who we collaborate with, who we cite, etc.

3. [5min] Overview: how to do a literature search?
   • What is a literature search? What is it for? Why is it hard?
   • Present a step-by-step guide on how to conduct a literature search
4. [15min] Small group activity #1: find the relevant keywords for a given research prompt (see example below)
5. [10min] Regroup: what were the keywords found? How did you know they were relevant? what was hard about this? (probably: knowing which keywords to search for, verifying a paper is relevant)
6. [5min] Presenting a guide to conduct a literature search
7. [20min] Small-group activity #2: given a research prompt and a list of papers: which are relevant? and in what ways are they relevant (i.e. which category do they belong to)? (see example below)
8. [20min] Regroup: go over list of papers and discuss
9. [5min] Go over paper tracking software (e.g. Zotero, the google scholar web extension), show example document keeping track of related work
10. [5min] In-class survey
    • Describe a new strategy you learned (if any) for doing a literature search for each of the following components: finding papers, reading papers, and tracking papers.
    • Is there anything else you took away? If so, what is it?

Insight: Students found it valuable to see how the instructors skim the paper abstracts, which keywords they focused on, which technical terms they glossed over, etc. to determine whether a paper is relevant.

Example Exercise

Given the research problem below (purposefully selected to be in a CS-adjacent field – statistical genetics – so that no student has prior knowledge):

1. Search for relevant keywords online. What are they? How did you know if they are relevant?
2. Given the list of papers below, which are relevant? In what way(s) are they relevant/irrelevant to the problem statement?

Problem: There is significant bias in the data collection used Genome-Wide Association Studies (GWAS) towards populations that identify as of European ancestry. As a result, statistical models of the data are substantially less predictive for populations identifying as of non-European ancestry.

Significance: Increased health disparities between populations identifying as of European vs. non-European ancestry.

Goal: To develop statistical methodology capable of both using existing data and effectively leveraging data from a more inclusive collection process to generalize well to populations identifying as of non-European ancestry.

Relevant: For each of the paper below, in what way it is (or is not) relevant to the problem statement? For example,

• The paper is irrelevant
• The paper motivates the problem
• The paper solves the same problem
• The paper solves a similar problem
• The paper solves a subproblem

List of papers:

1. Non-parametric genetic prediction of complex traits with latent Dirichlet process regression models
2. Current clinical use of polygenic scores will risk exacerbating health disparities Understanding the population structure correction regression
3. Understanding the population structure correction regression
4. Leveraging fine-mapping and non-European training data to improve trans-ethnic polygenic risk scores
5. Genes mirror geography within Europe
6. Improving Polygenic Prediction in Ancestrally Diverse Populations
7. Getting Genetic Ancestry Right for Science and Society
8. Multi-group Gaussian Processes
9. Human Demographic History Impacts Genetic Risk Prediction across Diverse Populations
10. Toward a fine-scale population health monitoring system