Should Machine Learning Be Trusted to Allocate Funding to Scientific Research?

E

x

p

l

o

r

a

t

i

v

e

p

a

r

t

What is the ethical dilemma?

Should machine learning be trusted to allocate funding to scientific research?

What is the content?

Nature Biotech has created a Dynamic Early-warning by Learning to Predict High Impact (DELPHI) system to predict which research^[a] will be the most ‘impactful’. Use cases include diversifying and providing security to funding portfolios.

What are the technologies or  types of data-usage involved?

Machine learning

What is the application? and  what drives the use in this  scenario/case/example

This framework is intended to find ‘diamond in the rough’ research that will have the greatest impact. With an issue as pressing as climate change, this tool could be incredibly valuable in selecting what research to fund.

Which ethical issues are at  play here?

This algorithm can be biased and make it even harder for people who already face challenges in getting their research funded.^[b] 

What group of people are at  risk? What group of people might gain?

Those who already face challenges getting their research funded will likely continue to be left out as existing biases in the scientific field are/ will be baked into the framework. Lenders wanting to take on less risk by funding research stand to gain by having a tool to select the most promising research to fund.

What is the wider impact of  this dilemma?

Climate change is a global issue but tools like this might focus the research to primarily impact/ benefit areas that hold the most capital to fund the research.

Cultural aspects important for this dilemma

Impact for who? If certain climate solutions work better in areas where less funding is available, will this research have a harder time getting funding from other sources?

S

o

l

u

t

i

o

n

o

r

i

e

n

t

e

d

p

a

r

t

Possible controls and  

comments on solution

Perhaps we can add some details on how machine learning is biased. For example, no matter how complicated the algorithm is, machine learning is based on data collected. That is to say, data itself may already be biased. “Entrenched discrimination — from the criminal legal system, to housing, to the workplace, to our financial systems. Bias is often baked into the outcomes the AI is asked to predict.”  Personally, I feel it is not the fault of machine learning. When we trace back to the starting point, we can see machine learning is like an oven, and it is the problem of ingredients of raw materials that matter?

Could this be acceptable if…?

Human oversight and/or expert interpretation before the data gets used

More transparency in the development of NLP or other algorithms for analyzing ESG strategies.

The data provided by the algorithms is viewed critically and not accepted as total authority on ESG strategy analysis

Any other observation and  conclusion