AI probably could have helped Google identify the ‘diversity memo’ sooner. This post looks at how AI could help other companies avoid these and many other complications around employee communication.
Last week a senior engineer, James Damore, at Google wrote a memo decrying Google’s diversity policy as illegal, and subsequently got himself fired. The memo sparked controversy over the weekend, after Damore argued that biological differences between men and women are the cause of the gender gap at Google and the tech industry as a whole. To read more on the incident and hear what others are saying about it you can check it out here, here or here.
In this blog post our intention is not to address the politically charged issue around Google’s now infamous diversity memo but rather to discuss tools which could help other companies avoid these and many other complications around employee communication. Whether it is internal memos on employee message boards or inappropriate posts on your various communication channels (i.e., Slack), the increasing level of connectedness in the workplace is a concern for all employers. Our goal is to use employee communication as a framework for exploring how AI can become an asset for your business.
We aren’t here to discuss whether the response was appropriate but instead…
…what AI could do to help
It is not possible for employers to manually supervise or inspect all inter-company communications let alone all communications from employees to external addresses on a given day. So how can you reduce the risk of a similar situation? We believe the solution is automated compliance monitoring using Artificial Intelligence (AI).
This solution consist of leveraging the power of artificial intelligence to keep an eye on communications on company infrastructure. While this may feel invasive or impermissible to some the use of AI to flag conversations, comments, or behaviours that are expressly forbidden means that some level of employee privacy can be respected while still allowing the employer to identify reprehensible behaviours. Permissible conversations are never viewed by humans, it is only when a communication crosses a defined line that a human inspects the conversation for follow on action.
On top of AI’s ability to consistently deliver strong results, the power of AI means that it can near instantly process millions of messages and flag situations for employers to follow up on. It is vital that companies have properly constructed and well thought out policy documents detailing what constitutes inappropriate communications internally and externally.
Could AI have helped Google identify the ‘diversity memo’ sooner? Probably. What this question exposes is some of the limitations that AI currently faces. In the absence of a clear policy the employees cannot be held accountable for their actions and the infraction cannot be identified. Once this type of infraction was identified could AI help Google identify harassment amongst employees in the future? Certainly.
AI is great at identifying patterns of behaviour and changes in behaviour. It is also extremely good at finding similar instances of bad behaviour. Giving employers the ability to provide examples of communications that they view as undesirable and having those communications flagged for follow up is a revolutionary and important tool in addressing workplace discrimination and harassment.
…how employers could leverage AI
Employer’s right now are either wilfully blind to the potential risk that harassment or discrimination in their organization poses or are forced to expend significant resources to remain vigilant. Leveraging AI to augment existing systems enables employers to take a more comprehensive approach than was previously available to them while protecting employees much more effectively.
AI expands the capacity of HR and management organizations to track and proactively address workplace harassment. Heading off these events at an early stage can represent significant long term savings for companies who implement these solutions. Instead of facing nasty litigation or suffering high employee turnover, AI can give you the tools necessary to head off toxic behaviour without requiring hours and hours of oversight.
…what employees should expect
Employees may expect some degree of privacy in the workplace. Whether they are right to do so exceeds the scope of this blog post. Employees should be made aware that using company infrastructure including email, text messaging, and network access means that they are acting on behalf of the company they work for. Misbehaviour online is no different than misbehaviour in the lunchroom – it should not be tolerated. Typically these policies are contained in the company’s acceptable usage policy.
Using AI to monitor company communications will inevitably be controversial amongst some employees. At the same time employees have already relinquished this privacy with the social networks that many employees frequent(Facebook, Instagram, Snapchat), the free email accounts they use (Gmail, Hotmail, Yahoo mail), and the free web browsers that deliver their content(Chrome, Firefox). Companies already monitor internet traffic of their employees – communication compliance is the logical next step.
Employees online communications are increasingly becoming too voluminous to monitor or check. Presently most companies do not have a clear strategy around data preservation and data monitoring. Miralaw has built a toolset that enables employers to monitor for traffic that they have identified as undesirable. The toolset flags communications that are similar in nature, theme or content to the undesirable examples and enables the administrator to review the communications. All communications are preserved and the administrator’s decisions are tracked and preserved for auditability purposes.
Current limitations of AI
AIs blindspots
AI is imperfect. While AI enables users to sift through millions of lines of text nearly instantly it also relies on being shown examples of all types of inappropriate communications before it is capable of detecting them. This means that it is theoretically a system that can be beat if users find new and creative ways that the administrator did not anticipate. Miralaw’s system uses state of the art to identify communications that have similar themes, structure and context to historically harassing content. The system learns from any new examples provided by the administrators and quickly adapts to identify new and historical communications matching the pattern identified.
Not all inappropriate communication is internal and not all inappropriate communication requires harassment. Many companies have communication policies with respect to external communications, communications with the press, or communications with customers and clients.
AI is useful at enforcing these types of communications standards as well. For example Company A may allow junior sales representatives to make pricing representations to customers but Company B does not. Company B would be able to flag quotes coming from junior sales representatives as inappropriate content and catch them before a communication was released contrary to policy.
At the same time if examples of these policies are not identified and shown to the system in advance they will be impossible to catch. It is for this reason that a company needs a well documented communications policy in advance of implementation of these systems.
Catching more diversity memos
It is unrealistic to assume that the diversity memo was the first communication by the engineer at Google expressing these feelings. It is equally unrealistic to assume that this engineer is the only person expressing these opinions in an organization that employs almost 62,000 people. It is very unrealistic for Google to manually monitor employee communications.
Using AI Google’s HR department could privately address the issue before it becomes a company wide disaster. This would undoubtedly have been the best path. Unfortunately Google did not have proper tooling or systems in place to head off these types of conflict.
Where to go next?
Compliance monitoring
We have been working hard to develop tools that enable comparative analysis of text. Our core products are built around processing large volumes of text to identify similarities and differences. We are experts at taking large volumes of text, processing them, and then providing valuable insights to our customers. We routinely do this in near realtime.
We have built compliance monitoring solutions for a variety of customers including an employment solution. These solutions enable customers to monitor and detect inappropriate communications using a hassle free interface.
Risk mitigation
Failing to implement solutions such as ours to monitor for compliance with communications and harassment policies may attract additional liability for employers in the future. We believe that a strong risk mitigation strategy leveraging the power of AI is the most sensible path forward. It is impractical to review each communication on an individual basis and, even if that were possible, it is not a good use of anyone’s time. Instead allowing algorithms to sift through the volumes of data and flag only the most important information means better use of everyone’s time.
We expect that tools like ours will increasingly become the norm and will become the standard against which a company’s efforts to avoid harassment will be evaluated.
Conclusion
Risk mitigation using AI may have helped Google intervene before a memo was published on a message board by a senior engineer helping them avoid the subsequent PR and HR disaster. Providing HR with the tools needed to facilitate early intervention in a cost effective manner is in every company’s best interest. Leveraging AI to track and monitor communications means that employees still receive a high degree of autonomy and privacy while still protecting the company’s best interests. Companies can avoid inappropriate communications using cutting edge technology and deliver results at a fraction of the cost of having humans manually reviewing documents.
Let us know what you think.
Compass is an innovative legal research platform exploring the frontiers of what is possible when delivering primary legal research content in new ways. We believe that with compass we are on the cutting edge of legal research by delivering results in a fashion similar to google as opposed to traditional research providers.
- Survival of the fittest: How can technology help small businesses thrive? - September 18, 2017
- Managing company communications with AI - August 15, 2017
