Can algorithmic decision-making help mitigate gender biases in Venture Capital decisions? New research from UVA Darden surely believes so.
Here’s a fact: In 2020, the total amount of Venture Capital (VC) picked up by women-led start-ups globally was about $4.9 billion (which was, in fact, a 27% decrease compared to the previous year: primarily owing to the reduced-investment atmosphere induced by the COVID-19 pandemic).
While this may not seem to be the worst performance given the massive hit most global investment took, given some perspective, it’s not the best either. The perspective: it accounts for only 3% of global VC investment and a dismal 2.3% of the total VC invested in start-ups. Concurrently, India alone saw venture capital investments worth around $10 billion — more than double the total VC investment received by women-led start-ups globally — in the same period.
Blame the Heuristics?
Why is this the case though – that start-ups founded by women ‘suffer’ from almost a systemic lack of investment? The answer, according to recent research by University of Virginia Darden School of Business’ Professor Morela Hernandez, is the fact that ‘VC investors will – often implicitly – prefer pitches made by men over those made by women.’ The cause, according to her, is often the heuristics based on past experiences or ‘rules of thumb’ that creep into funding decisions whilst considering risky investment perspectives.
According to Associate Director of Research at UVA Darden, Gosia Glinska: “When making decisions under a high degree of uncertainty,” says Hernandez, “investors will rely on heuristics for gauging potential success. The drawback of this approach is that these heuristics have been, in a sense, derived from limited data. Because past funding has gone to mostly male-founded start-ups, we lack data – and thus, heuristics – that integrate the female entrepreneur. It’s a huge impediment to realizing gender equity in VC funding decisions.”
The presence of this pervasive gender bias leads to several missed opportunities — and effectively leads to the widening of the gender gap: something that needs to be tackled head-on using machine learning technology, according to the researchers at Virginia.
Algorithmically-abled bias reduction?
To explore the role of algorithmic decision-making tools in reducing biases, Hernandez and fellow researcher Professor Roshni Raveendharan undertook a series of interviews of venture capitalists – both male and female – involved in both seed-stage and series-A investing. According to the Darden Blog: “At the same time, they analyzed VC firms’ use of data-driven decision-making tools, such as evidence-based forecasting models. Based on their findings, Hernandez and Raveendharan offer recommendations to help VCs mitigate bias in evaluating deals and making more quantitative – and equitable – investment decisions…”
From their study, the researchers found the presence of algorithmic aversion among several VC investors, i.e. the innate assumption that their human skills of assessing start up team dynamics and other relevant information through personal connections were more relevant to decision-making than using machine algorithms. To combat this, Hernandez and Raveendharan suggest several solutions, such as (i) enabling decision-makers to exert control/modify over the algorithmic decision-making process; and (ii) allowing for the presence of algorithmic advice instead of decisions, so that final judgement calls still remain in investors’ hands.
Furthermore, the researchers suggest several ways in which VC algorithms can be put to good use to tackle gender biases as well. This includes: (i) developing algorithms to tackle transparency issues and identify all potential scope for discrimination and (ii) releasing data on the performance and impacts of the algorithms instead of self-reported deal assessments from VC firms.
Whilst algorithmic decision-making for VC firms is still in its infancy, humans and algorithms together can mitigate several of the flaws that each have individually to produce more empirically-backed and less-biased outcomes.