AI based bias detectors – does my organisation really need them?

NLP for good

Bias is a reality.  An unwanted yet unavoidable side-effect of the wonder that is human cognition. An average human has close to 188 cognitive biases. The only way to stop these biases from transmogrifying into microaggressions and outright discrimination is personal awareness.  Yet because our biases are mostly unconscious, cultural and often institutional, surfacing them and accepting them is not always easy. And what’s worse is that the negative impact unconscious bias has on a person (hurt, shame, anger) is  disproportionately higher than the intention with which a statement is usually made. At the individual level, discrimination, including microaggression, can activate the body’s hormone and stress responses, potentially causing short- and long-term biological changes.

 

Businesses cannot afford to take the effects of unconscious biases lightly. The cost of managing incidents, employee morale, public perception of the brand, all gets affected. According to a survey by Gallup, actively disengaged employees cost the U.S. $450 billion to $550 billion in lost productivity per year. Furthermore, unconscious bias in hiring results in homogenous culture, which is highly averse to innovation and creativity. Diverse companies generally have  19% higher innovation revenue when compared to homogeneous companies. Similarly, ethnically diverse companies are 35% more likely to outperform, while gender-diverse companies are 15% more likely to do so. Plus, there is also the risk of losing out on amazing talent because  62% of people would go as far as turning down a job offer if it came from a culture that didn’t support a diverse workforce.

 

Although DEI trainings are crucial, it’s extremely difficult to change personal and cultural biases through short-term educational interventions. Here are a few limitations of DEI trainings.

 

  1. The positive effect wears down in a matter of a few days or even a few hours.
  2. There is a lack of standardisation within the training industry.
  3. They are often not personalised for how the biases can manifest in an individual’s personal style of communication.
  4. There is no clear way to measure or quantify the impact of these trainings

So it looks like an ideal DEI effort, which can truly make a difference in the long-term communicative behavior of people will need to be

  1. Consistent. 
  2. Standardised. 
  3. Measurable
  4. Personalised.
  5. Private
  6. Incentivized for users.

Enter AI based bias detectors! Whether it is  personal or collective unconscious bias or a genuine lack of awareness of the currently used inclusive terminology. inclusive language assistants can help people communicate without the fear of committing any inclusivity faux pas. These tools can even give you  intel on which words or phrases to use for the intended impact.

 

AI bias detectors can revolutionize how we communicate. Whether they be job descriptions, mails, websites, blogs or even personal communications, these tools encourage careful, thoughtful and most importantly, deliberate communication. Plus they meet all the six criteria mentioned above. They are research based, are continually updated and are customized for organizational policies. Most importantly, they are discrete and private to the individual user. Plus, the individual and the organization can even measure the progress and impact they are making.

 

Although AI based bias detectors have been on the market for sometime, there have been a lot of questions, concerns and misconceptions regarding their usefulness. Here are a few common misconceptions or “biases” that prevent customers from harnessing the power of such inclusivity tools for their DEI efforts.

 

Language, unless outrightly toxic, do not have any real world impact

Consider this “encouraging” statement you and I must have often made: “Women are as good as men in leadership roles” In the first read, this sentence seems to promote equality and inclusivity. But on careful inspection, we realize that  the statement subtly perpetuates the very stereotype it tries to debunk. The reason is that the sentence seems to build on the underlying, unquestioned premise “men are good in leadership roles”. Now compare this to “ “Women and men are equally good at leadership roles”. Here there is no underlying premise and hence is truly inclusive.  So if our positive and well-meaning statements can unconsciously propagate biases, without the sender or the receiver being none the wiser, imagine the stereotype havoc we must all be wreaking in our casual communications.

 

Also, how you frame your sentence and your choice of words can subconsciously impact how people perceive the same proposition. For instance,  a simple reframing of environmental issues from the narrative of harm, destruction and global warming  to that of values of purity — i.e. keeping our forests, drinking water, and skies pure , motivates more conservatives to support the cause.

 

So you see language matters and it can change our perceptions of the world. 

 

Changing the language merely treats the symptoms and not the root cause of discrimination. 

Yes, discrimination is a systemic problem, and  needs to be addressed at multiple levels. If not complemented by strong and consistent DEI effort, mere language changes alone may not yield results. But research shows that even slight word changes can create a ripple effect. Eg. Ads without male- or female-coded words such as ‘analytical’, ‘independent’, ‘compassionate’ etc. result in 41% lower Cost Per Application and 24% better apply-rate compared to ads with both gender coded words. 

 

The only real way forward for true diversity and inclusion is increasing representation at all levels in a company, and if the language we use is creating a barrier for talented people from underrepresented groups, changing it is perhaps the lowest hanging fruit for your DEI efforts that can be easily rectified. 

 

Inclusivity is just a passing trend

Wrong! By 2025, the workforce will be 75% millennials, 45% of whom actively seek out diversity in the workplace.By 2050, people of color will outnumber other groups, according to the U.S. Census Bureau. Nearly 40 percent of consumers come from multicultural backgrounds.  To future proof your market position your talent, your staff, your leaders must represent the market. Thus making diversity and inclusion the cornerstone of your organization is not a matter of choice any more.

 

Having a detailed inclusive language guide for communication is enough.

While it is great that your organization has an inclusive language guide, there is no way to ensure that it gets used at every instance, across the board. Also language is ever evolving and meaning is highly contextual. For example, calling a disabled person “differently abled” would have been considered fine a few years back; The word ‘negro’ although highly offensive in US, may not be deemed as offensive in other cultures. In many cases,  there could be an active unresolved debate going on  as to the correct way of referring to a community. Eg. whether to use “Latin”, “Latine” or “Latinx” 

 

Also, sometimes a word or a statement is biased only in a certain context. Eg. a word like “technological-savvy” is not biased per say, but may become one in the context of Job descriptions, wherein they have been found to be unappealing to older candidates.

 

When dealing with biases it is important to get these nuances right. A static piece of document, even if it gets revised may not have the same impact as a real-time solution backed by research. Plus you wouldn’t have any analytics as to patterns of bias and the measurement of impact.

 

This is too political and ‘woke’ for me

True inclusion also includes political inclusion. Workplace must be a safe place for all individuals regardless of world view or political affiliations. An essential workplace skill required in the 21st century is collaborating with people who do not share our views and may even be vehemently opposed to them. This is why DEI efforts must also have a personalized approach. Bias detectors should not be treated as arbiters telling you what you should or should not say but only as alerts. User discretion and decision are always paramount. 

 

The only purpose of bias detectors is to give a tip-off as to how a specific word you have used can be perceived and what impact it could potentially have, so that  every word, or sentence you use is intentional and serves your communicative intent well. These tools empower you to say exactly what you mean, eliminating scope for any misunderstanding, so that even though the general  political climate is polarized, your workplace does not have to be. 

 

Most of us want to be inclusive and positive with our communication. We want to do what’s right.  No one wants to offend or embarrass themselves and others intentionally. Let’s not get our unconscious biases stand in the way of our true intentions.   If what your organization needs  is an effective way of harnessing the inner workings of language for the larger good of inclusion, AI based bias detectors are indeed the future.

 

If you think your organization will benefit from AI bias detection, try assistive technologies  like Dost and DEI screener  from krita.ai that helps to eliminate all forms of bias and toxicity in all your communications, including job descriptions. Dost app is available on Slack, MS Teams and Lever platforms. 

 

About krita.ai

We help companies create bias-free communications at every touch point in the talent lifecycle. To do this, we built an AI assistant that is available on the most popular HR Tech software like Lever, Avature, SmartRecruiters and collaboration / communication platforms like Slack and MS teams. Our assistive technology provides real-time 1:1 feedback and inclusive language suggestions to people to make their communications inclusive of their target audience. 

 

Learn more about krita.ai here.

Comments are closed.

Thank you!

Click the button below to download the audit report.