Data sharing, research ethics, & scientific reproducibility

Rick Gilmore (

2017-09-28 11:29:37



  • Associate Professor of Psychology
  • Brain development, perceptual and cognitive development, big data, open science
  • Inaugural Director of Human Imaging, Social, Life, & Engineering Sciences Imaging Center (SLEIC)
  • Co-founder of the FIRSt Families database
  • Co-founder of the data library

Why should we care about data sharing?

Data sharing

  • easier than most PIs think
  • improves the reproducibility and robustness of research
  • maximizes benefits of participation in research

Data sharing…

  • can be accomplished for most types of data…
  • with advance planning and permission
  • increasingly mandatory for funders and journals
  • the responsibility of PIs & institutions

Data hoarding

  • is (sadly) the norm in some fields (e.g., psychology & neuroscience)
  • is problematic without very good reasons
  • undermines the reproducibility of research
  • wastes resources, time, & energy


  • The reproducibility “crisis” in science
  • The “crisis” in psychology
  • Challenges to more open data sharing & some solutions
  • Shared responsibilities

The reproducibility “crisis”

Is there a reproducibility crisis?

  • Yes, a significant crisis
  • Yes, a slight crisis
  • No crisis
  • Don’t know

Methods reproducibility

  • Enough details about materials & methods recorded (& reported)
  • Same results with same materials & methods

(Goodman et al., 2016)

Results reproducibility

  • Same results from independent study

(Goodman et al., 2016)

Inferential reproducibility

  • Same inferences from one or more studies or reanalyses

(Goodman et al., 2016)

Reproducibility crisis

  • Not just psychology
  • “Hard” sciences, too
  • Data collection to statistical analysis to reporting to publishing
  • Ethical urgency

Origins of the crisis in psychology & neuroscience

The sin of unreliability

Studies are underpowered

“Assuming a realistic range of prior probabilities for null hypotheses, false report probability is likely to exceed 50% for the whole literature.”

(Szucs & Ioannides, 2017)

The sin of hoarding…

The sin of corruptibility…

The sin of hurrying…

(Nuijten et al., 2015)

In our defense…

Psychology is harder than physics

Challenges to sharing

Belmont principles: Beneficence

  • Data sharing may pose risk of loss of privacy & confidentiality
  • But data sharing increases value, and participation value should be maximal

Belmont principles: Respect for persons

  • Data sharing may pose risk of unintended use of data
  • So participants should participate in decisionmaking

Belmont principles: Justice

  • Costs of research participation should be equitable
  • Benefits of research participation should be equitable

How to sharing data ethically

What data are you collecting?

  • Personally identifying or sensitive data?
  • What risks does data sharing pose?
  • How should data be protected?

Where will you share data?

  • Researcher/lab/university server
  • Supplemental material linked to journal article
  • Data repository

Where will you share data?

  • Researcher/lab/university server
  • Supplemental material linked to journal article
  • Data repository

Who will (& should) have access?

  • Public
  • Community of authorized individuals (researchers)
  • Individuals selected by repository
  • Individuals selected by data owner

Who will (& should) have access?

  • Public
  • Community of authorized individuals (researchers)
  • Individuals selected by repository
  • Individuals selected by data owner

What have participants been told, approved, understood?

  • What data collected, what will be shared
  • Who will have access
  • Where stored, how accessed
  • Purposes of use, types of questions
  • Shared for how long

Statutory, regulatory, or contractual restrictions?

Benefits (of sharing) vs. risks

  • How useful are data?
  • How sensitive are data?
  • How likely is it that reidentification could be achieved, and by whom?

Assessing disclosure risk

  • Reidentification by participants themselves
  • Reidentification by insider
  • Reidentification by targeted search (nemesis scenario)
  • Reidentification by mass matching (dystopian AI scenario)

Mitigating disclosure risk

  • Modify data
    • Aggregate or censor sensitive variables
    • Aggregate or censor secondary identifying variables
    • Perturb or add noise to variables

Case study:

Q: Does modification diminish value for secondary use?

A: “Defacing” MRI: No; “Defacing” video: Yes

Mitigating disclosure risk

  • Restrict access
    • Data enclaves (e.g., Census data)
    • Virtual data enclaves

Mitigating disclosure risk

  • Seek explicit sharing permission
  • Share unaltered with restricted audiences

  • Share identifiable data (video & audio recordings, individual metadata)
    • Unaltered to exploit inherent richness for reuse

  • Share across studies and researchers
    • Not on a study-by-study or PI-by-PI basis

  • Share consistently
    • Same permission structure
    • Same metadata
  • Store and serve data securely

  • Share identifiable data with permission
    • Template sharing permission language
    • ad hoc video/photo releases standard sharing levels

Seeking permission to share

  • Open sharing among authorized researchers, not public
    • Formal (institution-approved & signed) access agreement

  • Laws differ among countries
  • Comfort with data sharing varies among institutions
  • “Cloud” storage vs. institutionally housed
    • Store data (but not share) data for which no permission has been given?
  • Role of repositories
    • “Check” IRB/ethics training status or delegate responsibility to institutions

My opinion (Gilmore, Kennedy, & Adolph, 2017)

  • Biggest obstacle to more open sharing is cultural, not technical
  • Planning to share easier than post hoc
    • Everyone should be preparing to share
    • Share what data, share where, for how long, for what purposes?
  • Seeking permission to share should be common practice
    • Even if data aren’t sensitive and/or plan to de-identify
    • Can’t share (easily) if you don’t ask

Models for data sharing permission

  • Separate consent to participate from permission to share
    • Same protocol, two different forms
    • Two different protocols
  • Combine permission to share with consent to participate

Institutional roles


  • (Often) own data, although PIs don’t always know it
  • Have responsibilities under Federal guidelines to help PIs share data

“Investigators are expected to share with other researchers, at no more than incremental cost and within a reasonable time, the primary data, samples, physical collections and other supporting materials created or gathered in the course of work under NSF grants. Grantees are expected to encourage and facilitate such sharing.

– NSF, Dissemination and Sharing of Research Results,

Institutions should…

  • Partner with PIs
    • make planning for data sharing part of grant seeking & reporting process
    • ensure data & $ well-managed

Institutions should…

  • Make research quality part of criteria for promotion & advancement
  • Raise awareness that that reproducibility and research misconduct are genuine concerns
  • Strengthen research integrity offices, improve training programs, and develop programs to prevent research misconduct
  • Focus especially on mentoring and data management

Schrag & Purdy, 2017; Science

Compliance office/IRB role

  • Data destruction after X years sharing for limited indefinite period
  • Share data in lab website secure repository
  • Encourage seeking permission for narrow broad secondary use cases
  • Require Discourage reconsent of minors at age of majority unless data highly sensitive
  • Implement consistent treatment of similar data types
  • Encourage use of templates

Progress toward standard templates

In summary…

Data sharing

  • easier than most PIs think
  • improves the reproducibility and robustness of research
  • maximizes benefits of participation in research
  • increasingly mandatory for funders and journals

Data sharing

  • can be accomplished for most types of data with planning & permission
  • both PI and institutional responsibility
  • upholds scientific ideals
  • may be the most ethical thing to do (Brakewood & Poldrack, 2013)

Thank you


This talk was produced on 2017-09-28 in RStudio using R Markdown and the reveal.JS framework. The code and materials used to generate the slides may be found at Information about the R Session that produced the code is as follows:

## R version 3.4.1 (2017-06-30)
## Platform: x86_64-apple-darwin15.6.0 (64-bit)
## Running under: macOS Sierra 10.12.6
## Matrix products: default
## BLAS: /Library/Frameworks/R.framework/Versions/3.4/Resources/lib/libRblas.0.dylib
## LAPACK: /Library/Frameworks/R.framework/Versions/3.4/Resources/lib/libRlapack.dylib
## locale:
## [1] en_US.UTF-8/en_US.UTF-8/en_US.UTF-8/C/en_US.UTF-8/en_US.UTF-8
## attached base packages:
## [1] stats     graphics  grDevices utils     datasets  methods   base     
## other attached packages:
## [1] DiagrammeR_0.9.2 revealjs_0.9    
## loaded via a namespace (and not attached):
##  [1] Rcpp_0.12.12       compiler_3.4.1     RColorBrewer_1.1-2
##  [4] influenceR_0.1.0   plyr_1.8.4         bindr_0.1         
##  [7] viridis_0.4.0      tools_3.4.1        digest_0.6.12     
## [10] jsonlite_1.5       viridisLite_0.2.0  gtable_0.2.0      
## [13] evaluate_0.10.1    tibble_1.3.3       rgexf_0.15.3      
## [16] pkgconfig_2.0.1    rlang_0.1.2        igraph_1.1.2      
## [19] rstudioapi_0.6     yaml_2.1.14        bindrcpp_0.2      
## [22] gridExtra_2.2.1    downloader_0.4     dplyr_0.7.2       
## [25] stringr_1.2.0      knitr_1.17         htmlwidgets_0.9   
## [28] hms_0.3            grid_3.4.1         rprojroot_1.2     
## [31] glue_1.1.1         R6_2.2.2           Rook_1.1-1        
## [34] XML_3.98-1.9       rmarkdown_1.6      ggplot2_2.2.1     
## [37] tidyr_0.6.3        purrr_0.2.3        readr_1.1.1       
## [40] magrittr_1.5       backports_1.1.0    scales_0.5.0      
## [43] htmltools_0.3.6    assertthat_0.2.0   colorspace_1.3-2  
## [46] brew_1.0-6         stringi_1.1.5      visNetwork_2.0.1  
## [49] lazyeval_0.2.0     munsell_0.4.3