We design literature reviews across
the range of evidence needs
We are trusted leaders through publishing guidance on how to develop and communicate SLRs and advising on the optimal approach based on the evidence needed, including state-of-the-art AI assistance
Scoping
Understanding the evidence need and standards
Working within your constraints
Designing
Search strategy and methods
JCA- or PRISMA-compliant as needed
Reporting structure
Delivering
Quality-checked results
Reports that meet your needs
Collaborating
Working closely together
Managing time and budget
Our industry leading research in
AI in SLR
Artificial Intelligence for Rapid
Data Extraction
Data Extraction
AI can be used for data extraction, but we need to know how well it performs before we integrate into our workflows
We compared how Elicit AI performed against our
human experts when they extracted cost, humanistic
and study design data
With expert prompting, we saw strong performance
(78.0–95.0%) from the AI but a need for robust quality assurance protocols
We compared how Elicit AI performed against our
human experts when they extracted cost, humanistic
and study design data
With expert prompting, we saw strong performance
(78.0–95.0%) from the AI but a need for robust quality assurance protocols
Our industry leading research in
AI in SLR
An AI-Enhanced Targeted Literature Review Workflow to Support Inclusive Research Practices and Population Science Research
We evaluated the performance of a GPT-4-assisted title
and abstract screening workflow for supporting TLRs
with high screening burdens
Sensitivity and specificity were improved with systematic prompt refinement, reducing the number of missed relevant articles while enhancing overall accuracy,
but human oversight remained essential
and abstract screening workflow for supporting TLRs
with high screening burdens
Sensitivity and specificity were improved with systematic prompt refinement, reducing the number of missed relevant articles while enhancing overall accuracy,
but human oversight remained essential
Our industry leading research in
AI in SLR
Bigger, Better Studies
Our new approach allows us to rapidly find the most relevant studies, based on a reproducible Boolean search
We used AI to extract simple data at title/abstract screening and then ranked the studies
This allows us to adjust sensitivity and rapidly prioritize the most relevant studies, resulting in reports that include the bigger, better studies to rapidly give answers to questions and inform next steps
We used AI to extract simple data at title/abstract screening and then ranked the studies
This allows us to adjust sensitivity and rapidly prioritize the most relevant studies, resulting in reports that include the bigger, better studies to rapidly give answers to questions and inform next steps
Our industry leading research in
AI in SLR
Testing Automated Prompt Engineering Strategies for Systematic Literature Review Screening
We tested different techniques for how we can engineer
the best prompts for title/abstract screening
For each, we calculated how accurately the prompts
identified or excluded studies and found big differences between strategies
We are using these results to choose the best approach
based on the needs of each review
the best prompts for title/abstract screening
For each, we calculated how accurately the prompts
identified or excluded studies and found big differences between strategies
We are using these results to choose the best approach
based on the needs of each review
Our industry leading research in
AI in SLR
Artificial Intelligence for Rapid
Data Extraction
Data Extraction
AI can be used for data extraction, but we need to know how well it performs before we integrate into our workflows
We compared how Elicit AI performed against our human experts when they extracted cost, humanistic and study design data
With expert prompting, we saw strong performance (78.0–95.0%) from the AI but a need for robust quality assurance protocols
We compared how Elicit AI performed against our human experts when they extracted cost, humanistic and study design data
With expert prompting, we saw strong performance (78.0–95.0%) from the AI but a need for robust quality assurance protocols
Our industry leading research in
AI in SLR
An AI-Enhanced Targeted Literature Review Workflow to Support Inclusive Research Practices and Population Science Research
We evaluated the performance of a GPT-4-assisted title and abstract screening workflow for supporting TLRs with high screening burdens
Sensitivity and specificity were improved with systematic prompt refinement, reducing the number of missed relevant articles while enhancing overall accuracy, but human oversight remained essential
Sensitivity and specificity were improved with systematic prompt refinement, reducing the number of missed relevant articles while enhancing overall accuracy, but human oversight remained essential
Our industry leading research in
AI in SLR
Bigger, Better Studies
relevant studies, based on a reproducible Boolean search
We used AI to extract simple data at title/abstract screening and then ranked the studies
This allows us to adjust sensitivity and rapidly prioritize the most relevant studies, resulting in reports that include the bigger, better studies to rapidly give answers to questions and inform next steps
We used AI to extract simple data at title/abstract screening and then ranked the studies
This allows us to adjust sensitivity and rapidly prioritize the most relevant studies, resulting in reports that include the bigger, better studies to rapidly give answers to questions and inform next steps
Our industry leading research in
AI in SLR
Testing Automated Prompt Engineering Strategies for Systematic Literature Review Screening
We tested different techniques for how we can engineer the best prompts for title/abstract screening
For each, we calculated how accurately the prompts identified or excluded studies and found big differences between strategies
We are using these results to choose the best approach based on the needs of each review
For each, we calculated how accurately the prompts identified or excluded studies and found big differences between strategies
We are using these results to choose the best approach based on the needs of each review
Connect with our experts
Get in touch to learn how we combine AI with scientific rigour for impactful results
Gemma Carter
Communications Director
gemma.carter@pharmagenesis.com
Polly Field
Communications Director
polly.field@pharmagenesis.com
Christian Eichinger
Scientific Director
christian@pharmagenesis.com
