While data and analytics, and the technologies that render them explicit, play an important enabling role in improving diversity, equity and inclusion in organizations and society, algorithmic methods such as artificial intelligence (AI) and machine learning (ML) bring potential risks that must be recognized and addressed. It is incumbent upon strategists, policy makers, vendors, implementers and users of AI/ML systems to take the appropriate steps to safeguard individuals from potential harmful outcomes these technologies can deliver. It’s also important to recognize that data, analytics and technology are necessary but not sufficient for achieving the increasingly diverse, inclusive and equitable organizations that many are striving for; to create a better future, bold, programmatic actions are needed to achieve the outcomes to which many of us aspire.
Recognizing and mitigating risks
Much has been written and claimed about benefits that AI/ML can deliver in a multitude of contexts and domains. As a result of some of those claims, unrealistic expectations abound regarding presumed accuracy and simplicity of algorithm-based applications. But the reality is, getting desired results from analytics and technology requires really hard work. This is not simply “plug and play” technology, and users – including HR practitioners – must be hyper-vigilant about deeply understanding our datasets and their limitations; who’s involved (and who’s excluded) in sourcing and curating data and building analytical models; what results are generated by the models; how the models are deployed and who’s potentially harmed by them; and the full spectrum of outcomes they can produce. This is the hard work that’s necessary if we want data and technology to contribute positively to equity and inclusion, and not to perpetuate historical biases and injustices at scale.
“Doing the work” to deliver the desired outcomes
All stakeholders have a role to play in doing the work required for producing positive outcomes and mitigating negative repercussions: government entities, educational institutions, private corporations, buyers and users of these technologies, and individuals whose data are being consumed and who are impacted by the model outputs. As an influential corporate entity in the technology industry, IBM has a long history of advocating for and contributing to the development of socially responsible technology. In addition to establishing its own internal AI Ethics Board and making available numerous open-source resources supporting trustworthy and explainable AI, the company has made many contributions to the development of government legislation and standards. Among these contributions are recently published recommended guidelines for government policy to help minimize instances of bias in AI systems (https://www.ibm.com/policy/mitigating-ai-bias/).
Standards and policies are important mechanisms for preventing harm and abuse. History would suggest that organizations (and individuals) are not very good at regulating themselves in the absence of clearly agreed rules and guidelines, so governing entities play a key role in steering the use of data and technology in the right direction for individual and societal wellbeing. Governmental guidelines also serve to level the playing field, such that all organizations are operating by a consistent set of rules.
The recommended guidelines from IBM focus on five policy priorities:
- Strengthen AI Literacy Throughout Society. As AI systems become more pervasive and influential, it’s incumbent upon all of us to understand these systems, how they impact our lives, and how to engage with them. Governments can help by ensuring the appropriate education is incorporated into school curricula, as well as providing funding to increase diversity across the AI ecosystem, from developers to users of the systems.
- Require Assessments & Testing for High-Risk AI Systems. Systems that are deemed “high risk,” where consequences for individuals are potentially substantial (such as law enforcement use cases), continual, robust, and transparent bias testing and mitigation are called for, based on agreed standards. Documentation must be auditable, and governments should provide enablement to the developer community to support their compliance.
- Require AI Transparency Through Disclosure. Users of these systems should clearly understand when they are interacting with technology vs. a human being, and they should be informed when AI is used to make high-risk decisions about them. This includes disclosing “why” as well as “how” the decisions are made.
- Require Mechanisms for Consumer Insight and Feedback. When used for making high-risk decisions, individuals should have a means of providing feedback based on those decisions, as well as mechanisms for issuing complaints. System owners should be required to review users’ concerns and to address systemic issues.
- Establish Universal Use Limitations of AI and Adopt Responsible Licensing Practices. Guardrails should be put in place to prevent harmful use of AI systems, by establishing universal limitations (e.g., not allowing racial profiling or mass surveillance). Licensing clauses should be used by developers to limit the use of potentially harmful systems.
Universal adoption of guiding principles such as these can limit harm while allowing innovation to flourish.
Recognizing the limitations of algorithms in building a better future
Data, analytics and technology bring needed visibility to past and present data, and the proper guardrails can protect users from risks associated with algorithmic prediction and decision-making. But what if we want (and need) to create a future that differs substantially from the past? This is not what algorithms are designed to do; rather, they operate precisely by predicting the future from the past, by identifying patterns in historical data to make future predictions. To create a different, more inclusive and equitable future, we need to actively recognize and remove systemic barriers and constricting environments that are deeply entrenched in society and standing in the way of what we’re trying to achieve.
Creating new pathways to opportunity
Amid ongoing and growing perceptions of labor shortages and skill gaps, an abundance of talented people exist who have historically been subject to systemic barriers preventing them from realizing opportunities and fulfilling the demand with their capabilities. Rather than continuing to operate within these ingrained societal constraints, what if we could create new pathways to opportunities? A number of promising examples exist of bold actions to create such pathways, which have the dual benefit of filling skill demand while simultaneously building more diverse workforces.
IBM’s actions to create new pathways
IBM has a long history of implementing opportunity-creating actions, policies and advocacy, and the events of the past year have reinforced the importance and value of bold, game-changing actions. A few examples are highlighted below.
Tech Re-Entry: A path back to technical careers
IBM recognizes there are many reasons people need to take a break in their careers, and that doing so can make it difficult to re-enter the workforce, especially considering how quickly technology and jobs evolve today. To address that challenge and create a pathway for talented people to rejoin the workforce after a break, IBM created the Tech Re-Entry Program. This is a flexible and paid program to re-enter the workforce, with a path to full-time employment. Members of the program are enabled with mentors, and they participate as a cohort that experiences the program together and provides mutual support throughout.
Apprenticeship Program: Building technical skills while getting paid
Another noteworthy pathway is the Apprenticeship program, which allows people without advanced degrees to build new technical skills, opening doors to a new career. Similar to the Tech Re-Entry program, participants in the Apprenticeship program receive a salary. This is an essential defining feature that makes this is a viable option for people of all socioeconomic levels, essentially leveling the playing field by providing an equitable opportunity for all.
P-TECH: An innovative approach to public education
Pathways in Technology Early College High School, or P-TECH, is a path to a no-cost, skills-based associates degree. This is an innovative public school model spanning grades 9 to “14,” bringing together the best elements of high school, college and career. Students graduate with a no-cost, industry-recognized associates degree that will enable them to secure a competitive entry-level position in a growing science / technology / engineering / mathematics (STEM) field, or to continue and complete study in a four-year higher education institution. The model was co-developed with IBM and professional educators, policymakers and elected officials. It’s designed to be both widely replicable and sustainable, as part of a national effort to reform career and education. There are now 200 P-TECH schools with more than 100,000 students in 18 countries, and P-TECH students are supported by more than 500 business partners (of which IBM is one).
OneTen: a pathway for Black talent
OneTen is a consortium of companies that has committed to hiring one million Black workers into family-sustaining jobs within 10 years. IBM is a co-founding member, with former CEO Ginni Rometty a co-leader of the initiative. These are jobs that pay a living wage, are at low risk of being replaced by automation, and do not require a 4-year degree or excessive years of prior experience. And, the OneTen commitment goes beyond hiring: the CEOs of the member companies have committed to retaining, developing and advancing these workers into meaningful careers with their companies. OneTen is designed as an ecosystem consisting of (1) employers (the member companies), committed to creating opportunities and inclusive cultures; (2) talent developers, providing education and skill-building for diverse talent; and (3) talented Black individuals looking for career opportunities. While OneTen is in the early stages of establishing the infrastructure and logistics to support the program, this is a promising example of the bold action that’s needed to truly accelerate diversity, equity and inclusion.
The indispensable role for data, analytics and technology
Given these exciting, game-changing actions for creating more diverse, equitable and inclusive organizations, is there a role for data, analytics and technology in creating that future? Yes, absolutely. First, it’s necessary to have a detailed understanding of the current state of diversity, equity and inclusion in an organization, in order to identify where the problems are, the magnitude of those problems, and where to focus efforts to address the issues. This is only possible through close examination of quantitative and qualitative data, rendered explicit and analyzed with the appropriate tools and technologies. Second, it’s important to analyze the effectiveness of actions and interventions, to understand what works and what doesn’t and to properly focus efforts on solutions that yield the best results. Observations and conclusions must be grounded in evidence supported by data, ideally collected and analyzed consistent with the scientific method using rigorous experimental design (reliable and valid pre/post measures, random assignment to experimental and control groups, sufficient sample sizes, etc.). Clearly, data, analytics and technology play an indispensable role in improving diversity, equity and inclusion, as essential supporting mechanisms.
The right tools, appropriate safeguards, and bold action
Data, analytics and technology did not create the diversity, equity and inclusion challenges we struggle with in organizations and society today, and they will not, on their own, solve them. Rather, bold, programmatic actions and initiatives are needed to move beyond the constraints of the past. Examples of such opportunity-creating initiatives have emerged over the past decade, with the promise of eroding long-established barriers and enabling success. Data, analytics and technology play an essential supporting role. Be mindful and intentional about where analytics bring the most value. Remain cognizant and vigilant in recognizing inappropriate applications of AI/ML systems and mitigating any potential harm they might cause. Support and adhere to guardrails that prevent algorithmic harm and abuse. Embrace bold actions and initiatives for removing ingrained barriers impacting groups and individuals. And systematically analyze the efficacy of initiatives with well-designed research studies that provide data-based evidence of what works, what doesn’t, and what’s required for success. This will ensure efforts result in the diverse, equitable and inclusive workplaces to which we aspire, creating a future that’s different and better than the past.