MVP vs EVP: Ethics in Agile Startups

The Evolving Startup Model: Beyond the MVP

The typical growth path of a startup is widely understood: conceive an idea, assemble a team, and quickly create a minimum viable product (MVP) for user testing.

However, contemporary startups must re-evaluate the traditional MVP approach. The increasing prevalence of artificial intelligence (AI) and machine learning (ML) in technology, coupled with growing public awareness of the ethical considerations surrounding AI’s impact on human decision-making, necessitates a shift.

The Power of the MVP and its Limitations

An MVP facilitates the collection of vital feedback from the intended audience. This feedback then guides the minimal development needed for product launch, establishing a robust feedback loop central to today’s customer-centric businesses.

This lean and agile methodology has proven remarkably successful over the last twenty years, fostering the creation of countless startups, some achieving valuations exceeding a billion dollars.

But simply building products that function effectively for the majority is no longer sufficient.

The Rise of Ethical Concerns in AI

Recent years have witnessed the discontinuation of several AI- or ML-driven products due to ethical issues surfacing after substantial investment. Examples include facial recognition technology exhibiting bias against certain demographics and credit-lending algorithms demonstrating discriminatory practices.

In today’s competitive landscape, where a single opportunity can determine success or failure, this risk can be devastating, even for established organizations.

Introducing the Ethically Viable Product (EVP)

Startups need not abandon the lean business model entirely. A balanced approach exists, integrating ethics into the startup mindset without compromising agility.

This begins with redefining the initial objective – demonstrating an early-stage proof of concept to potential customers.

Instead of focusing solely on an MVP, companies should prioritize developing and launching an ethically viable product (EVP). This is built upon the principles of responsible artificial intelligence (RAI).

RAI encompasses a thorough consideration of ethical, moral, legal, cultural, sustainable, and socio-economic factors throughout the AI/ML system’s lifecycle – from development and deployment to ongoing use.

This practice isn’t solely beneficial for startups; it also represents a valuable standard for larger technology companies engaged in AI/ML product development.

Three Steps to Developing an EVP

Here are three actionable steps startups – particularly those heavily reliant on AI/ML – can take to create an EVP:

• Consider the ethical implications from the outset.
• Prioritize responsible AI principles during development.
• Integrate ethical considerations into the product lifecycle.

Establishing Ethical Leadership: The Role of a Chief Ethics Officer

Many startups prioritize roles like Chief Strategy Officer or Chief Investment Officer. However, a Chief Ethics Officer is equally, if not more, crucial for long-term success.

This individual serves as a central point of contact, ensuring the startup’s product development aligns with established ethical guidelines from the company itself, the broader market, and public expectations.

Facilitating Ethical Alignment Across Teams

The Chief Ethics Officer functions as a key communicator. They bridge the gap between founders, the C-suite, investors, the board of directors, and the development team.

Their primary responsibility is to foster a culture of proactive ethical consideration and risk mitigation throughout the organization.

The Persistence of Bias in AI Systems

Artificial intelligence systems learn from existing data. Consequently, if current business practices contain systemic biases – such as disparities in lending based on race or gender – these biases will be replicated and perpetuated by the AI.

Simply replacing the data after discovering an ethical issue isn't a viable solution. The algorithms have already been trained, and that initial influence is lasting.

The Irreversible Impact of Early Training

The impact of initial data on an AI’s behavior is akin to the formative influences of childhood. Just as an individual cannot erase the impact of their upbringing, an AI cannot fully overcome its initial training.

Therefore, Chief Ethics Officers must proactively identify and address inherent biases within the organization before they become embedded in AI-driven products. This requires diligent oversight and a commitment to ethical development practices.

Embedding Ethical Considerations Throughout AI Development

Truly responsible Artificial Intelligence isn't a one-time check; it’s a comprehensive governance structure. This framework centers on identifying and mitigating the risks inherent in an organization’s entire AI lifecycle. Consequently, ethical considerations must be woven into every stage of development – from initial strategy and planning, through to deployment and ongoing operations.

When defining the project scope, the development team should collaborate with the chief ethics officer. This collaboration ensures awareness of broad ethical AI principles. These principles establish behavioral guidelines applicable across diverse cultural and geographical contexts.

A thorough risk and harm assessment is paramount. This assessment must pinpoint potential threats to individuals’ physical, emotional, and financial security. Furthermore, it should evaluate the potential environmental impact of the AI solution, considering sustainability.

Throughout the development process, continuous evaluation is crucial. Teams should consistently assess whether their AI applications align with the company’s core values.

Specifically, they must verify that models demonstrate fairness across all demographics and uphold individuals’ privacy rights. Robustness, security, and the effectiveness of the operational model in ensuring accountability and quality should also be regularly examined.

The data used for training is a fundamental aspect of any machine learning model. Organizations should focus not only on the initial Minimum Viable Product (MVP) and proof of concept, but also on the potential future scale and geographic distribution of the model.

This foresight enables the selection of a truly representative dataset, proactively preventing potential issues related to data bias. Selecting appropriate data is vital for long-term, ethical AI performance.

The Importance of Continuous AI Governance and Regulatory Adherence

Considering the broad societal effects, it is anticipated that the European Union, the United States, or another governing body will enact consumer protection legislation concerning the application of AI/ML. Following enactment, these safeguards will likely extend to other global regions and markets.

This pattern has been observed previously: the implementation of the General Data Protection Regulation (GDPR) within the EU spurred a global surge in consumer protections. These new rules necessitate that organizations demonstrate explicit consent for the collection of personal data. Currently, stakeholders across the political and commercial spheres are advocating for ethical standards surrounding artificial intelligence.

Startups utilizing AI/ML-powered products or services should proactively prepare to showcase ongoing governance and adherence to regulations. Integrating these procedures now, before regulations are mandated, is a crucial element of Effective Validation Planning (EVP).

Furthermore, a review of the regulatory and policy environment is recommended before product launch. Including an individual with direct involvement in current global discussions on AI within your board of directors or advisory board can provide valuable insight into potential future developments. Regulatory changes are inevitable, and preparedness is key.

The Benefits and Potential Risks of AI/ML

The potential advantages of AI/ML for humanity are undeniable. The capacity to automate repetitive tasks, optimize business workflows, and enhance customer interactions is substantial. However, startups must acknowledge the potential impacts of AI/ML on their customers, the broader market, and society as a whole.

For startups, success often hinges on a single opportunity. It would be detrimental if a promising product were to fail due to ethical issues that were not identified until after its release. Therefore, startups must embed ethical considerations into the development process from its inception, establish an EVP grounded in Responsible AI (RAI), and maintain robust AI governance even after launch.

AI represents the future of commerce, but it is vital to maintain a focus on empathy and the human aspect of innovation.