Data & AI

Technology and Telecommunications

Part 3: When and where is AI ethics needed?

It's nice to know that AI ethics is a must when your organization is deploying AI to solve societal problems or better serve the public. Simply putting in the money won't take you very far. Making AI work for society is the work of people after all. But how to do it then? When and where should ethical reflection on AI be introduced in public organizations?  
In this final part of the introduction series on "AI ethics for public organizations," the separate phases of AI usage are distinguished. They are related to helpful ethical questions that are specific to those phases, and we share insights on where to situate the AI ethics capability in your organization.

Thomas Mollema

Netherlands

Digital

Mitigate risks and harness potential

Part 2: AI ethics

Part 2: AI ethics: mitigate risks and harness potential

If you want to apply AI to your organization's problems, where do you start? How do you do so in line with the public values you want to protect? Like all technologies, whether AI helps or harms depends on how you use it. But ignorant use of AI is more likely to do harm than do good. This means not reflecting on how AI relates to your organization's goals and values. 'Why is AI ethics necessary for every organization using AI systems?' is therefore the second question this series of introductions on "AI ethics for public organizations" addresses.

Part 1: The basics of AI ethics

‘What really is the ethics of AI?’ We often get this question from public organizations. Joining AI and ethics might seem strange at first glance, but raising ethical questions is essential to make AI work for your organization. No automating technology is a 'quick fix' for the difficult societal problems that public organizations deal with. Only AI systems that are <a href="https://highberg.com/insights/truly-valuable-services-are-sensible-from-the-initial-design" target="_blank" rel="noreferrer noopener">'sensible by design'</a>, well thought-out, and steer clear of ethical risks succeed in helping solve large, societal challenges. Think of making health care more effective, efficiently sharing information with stakeholders, and strengthening social security and well-being.  
The role of ethics in applying AI to societal challenges is therefore worth examining.

Pride Comes Before a Fall

Is Generative AI the Digital Tower of Babel?

Imagine a mysterious disease ravaging a remote tribe in Papua New Guinea. The brains of those affected degenerate rapidly, and scientists soon discover the cause—a ritual in which the dead are consumed. This disease, later known as Kuru, spreads as the tribe literally ate themselves.
Now, decades later, we see a similar phenomenon in the world of artificial intelligence (AI). No, our computers haven’t started eating each other (thank goodness!), but something disturbing is happening: Generative AI systems (hereinafter simply AI for readability) are increasingly “consuming” their own output, with potentially disastrous consequences.
Welcome to the world of 'model collapse' and 'shitification' - two-dollar words for a multi-billion dollar problem.

Erik de Ruiter

Is Generative AI the Digital Tower of Babel? Pride Comes Before a Fall

AI ethics for public organization: A introduction series

AI ethics for public organizations

AI ethics for public organizations: A primer series

In the public debate, artificial intelligence (AI) is seen as means to increase worker productivity and provide better and more health care. AI could also speed up and automate administrative processes, distribute energy and resources more efficiently, and better detect natural hazards. AI is simultaneously a technology and a social promise. Public organizations are raising questions about AI. Should they do something with AI? How does AI relate to the problems that they must solve?

AI ethics for public organizations: A introduction series

Organizational

Even without a plan, GenAI finds its way into your organisation

GenAI finds its way into your organisation

‘Our employees know the risks of ChatGPT’ is a statement we hear in our daily work, but does this ring true? There are plenty of examples of generative AI (genAI for short) that have been deployed in less than optimal ways. That is why it sometimes seems better not to use it at all. Advisors and users think that deploying genAI will save costs and that you will escape all the risks everyone orates about on LinkedIn. In this blog, we offer practical guidance on how to get started with a genAI strategy.

Jules van den berg

Daan Smits

Two pager: sustainably implementing Data ethics

Two pager: Data ethics

Joshua sortino Lq Khn Dz SF 8 unsplash

In recent years, data ethics has become an important topic when utilizing data and data applications. Responsible data usage is often taken as a starting point. However, in practice, sustainably implementing data ethics within organizations proves to be challenging. Over the past few years, Highberg has undertaken various projects related to data ethics. Our experience has shown that data ethics is closely linked to organizational change. Read our two pager and let us know what you think!

Marlijn Mulder

Workshop AI for HR

Do you recognize yourself in the following situation?
You're not sure if AI is just a hype, but you also feel like you should do something with it because if it’s not a hype, you don't want to miss the AI boat.
You've read a bit about AI, but it all remains very abstract. You need an explanation in simple, straightforward language.
You want to practice applying AI tools to HR topics such as employer branding, recruitment, diversity &amp; inclusion, engagement, and employment conditions in a playful and safe way.
In that case, the practical workshop "AI for HR" is something for you and your team!

Risk, Resilience & Compliance

Government and Public Sector

Most important obligations in the AI Act: how to comply now?

Break through in regulation of AI: The AI Act

This text is translated from Dutch with the help of AI - On December 8, 2023, the European Union reached a preliminary political agreement on the Artificial Intelligence Regulation (hereafter referred to as the AI Act). The final text is formal approved by the European Parliament. 20 days after official publication the AI Act will enter into force. This landmark legislation marks the EU's first step towards regulating artificial intelligence, targeting developers, distributors, and users of AI systems, including those systems used for predictions, classifications, and analyses.
The potential impact on governments and businesses deploying AI systems is significant. In this insight, we'll walk you through the key obligations, how they affect your organization, and proactive steps you can take to ensure compliance.

Sabine Steenwinkel-den Daas

Germany

France

Responsible AI: how to achieve AI fairness with understanding and mitigation bias by AI

Bias in AI Explained

This text is translated from Dutch with the help of AI - "Tell me honestly, what's your biggest fear when it comes to algorithms and AI?" When I ask people this question, the response often is, "AI-systems discriminate!" And of course, discriminating is far from desirable. But how do we ensure that algorithms don't discriminate?
First and foremost, it's important to understand that algorithms themselves do not discriminate. However, algorithms can be developed in ways that lead to unwanted or undesired distinctions in representation of groups in the result. This is commonly referred to as "bias" in technical terms. 
Essentially, if you have control over the bias within the algorithm, you also minimalize discrimination. In this insight, I will explain how to achieve this. Let's start by answering the question: What is bias?

Control AI risks and make impact: top criteria for responsible deployment of AI systems

In control of your AI systems

This text is translated from Dutch with the help of AI - Have you ever heard or said within your organization, "We're working with data/algorithms/AI" or "We have plenty of data initiatives, but scaling within the organization is challenging"? These questions all boil down to one main question: How do you ensure that you apply the right algorithms in the right way?

When is transparancy on AI sufficient? Comply to the AI Act with an algorithm registry?

Transparency with algorithm registry?

This article has been translated from Dutch with the help of AI: In an era where algorithms exert increasing influence on our daily lives, maintaining control over the latest developments and being transparent about them is more important than ever. This leads us to a crucial question: when is the openness and transparency surrounding these algorithms sufficient? The answer is not simple, because transparency goes beyond just registering the algorithm. In this article, we delve deeper into the steps needed to start working on transparent and responsible use of algorithms at 5 maturity levels.

Small college perform better Learning analytics offers opportunities for large institution

Scale affects study yield

Uit de data blijkt schaal is ook in hoger beroepsonderwijs van invloed op studierendement

Data shows: scale also influences study efficiency in higher vocational education   

What do the data tell us about the quality of education and study efficiency in higher vocational education? And what tools do the data offer for improving this? We have collected, analyzed and visualized open data on higher vocational education in a dashboard. This dashboard can be found on our website.   

From the analysis it can be seen that the study efficiency after 5 years of education in HBO has decreased by about 10 percentage points since 2002. Whereas almost 60% of the students who started their studies in 2002 (the so-called 2002 cohort) graduated in 5 years, this is only about 50% for students who started 10 years later. 
For universities, the smaller ones do better than the larger ones. This is also true for colleges of higher education. The dashboard shows that large institutions (with more than 20,000 students) score below the national average (the black line in the figure) on study efficiency. However, the gap with the national average has narrowed substantially since the 2000 cohort. 
To better understand the relationship between student success and institution size, we performed a correlation analysis. 
This clearly shows that the small institutions are doing better than the large institutions: the small institutions are in the top left of the figure and the large ones in the bottom right. 
One explanation for the better performance of small institutions may be that they offer very specific degree programs (especially in the arts, agri-business and based on a specific foundation/signature). 
Students who choose these institutions are thus making a more conscious choice for a specific program. In academic education, this can be seen at Wageningen University. This university has come out on top in the Study Choice Guide every year since 2004. Students who choose this university in "provincial village" Wageningen must be "very motivated," according to the guide.

Gijs de Groot

Data shows: scale also affects study efficiency in higher vocational education

Student characteristics predict academic success. Customized support possible

Chracteristics as study success predictors

Learning Analytics Kenmerken van studenten als voorspellers voor studiesucces

In the previous blogs, we explained what learning analytics is. We also used a dashboard to show how scale in higher vocational education affects study success and what interventions are possible based on this information. However, scale of an educational institution is not the only variable that affects study success. After all, students differ from one another. Are there certain types, or profiles, of students who do better than others? In this third and final blog of this series, we will look for an answer to this question, again using the dashboard.

Learning analytics: student characteristics as predictors of study success

Centralized, decentralized or hub and spoke? Efficient collaboration, clear structure.

The hub and the spokes

De naaf en de spaken hoe organiseren we onze data expertise

How do we get more out of our data? We ask our data specialists, but where are they in our organization? In a separate corner behind big screens and fat computers, among employees in departments or somewhere in between?

The hub and the spokes: how do we organize our data expertise?

Effective visualizations: Audience, setting, omitting, imagery, clarity, color.

This is how to deal with visualization urges

Zo ga je om met visualisatiedrang 6 tips voor de vormgeving van visualisaties

Because visualizations within organizations can be created in many ways, the search for the right form is on. Design choices affect the outcome. The content may be seemingly identical, while a powerful design ensures that the message is better understood or comes across better. The following 6 tips will help in the choice of design. 

Tip 1: Choose a clear target audience for the visualization. 
A director does not want complex Excel sheets, while others want to study the details. Therefore, match the message to the target audience. If the target group is organization-wide make sure it is accessible to everyone and keep it as simple as possible. 

Tip 2: Choose the setting in which the visualization should land. 
Is it a report from tooling, the report and presentation of a study, a communication message, a drawing on whiteboard/flip chart, a supplement to a quotation, etc., the setting gives direction to the design. A communication message should only be about the message, while a visualization in a quotation should be visually appealing and also support content. 

Tip 3: The art of omission. 
Ask whether it is really necessary to present certain information. The more information is left out the more focus there is for what is left. This makes a discussion much more specific and gets people thinking in a more focused way. 

Tip 4: Within visualizations, use figures instead of text. 
If, within a visualization, a word or phrase can be replaced with a figure, it is easier to understand and sticks better. This works the opposite way if the chosen figure is not sufficiently clear, then discussion and ambiguity will arise. 

Tip 5: Provide overview within the field of view. 
For example, an infographic fits into one page, but if there is a lot on it, the message is lost. On the other hand, if there are more visualizations for the same topic, ask yourself if it can't be combined in clickable selections, for example. 

Tip 6: Use lots of color. 
Dark colors are more likely to be boring. Bright colors appeal, but are not always easy to see or it is childish. Therefore, when adding color, always make sure it is the right shape and accessible. 

Extra tip: Contact me to go over even more considerations for creating powerful visualizations.

Douwe Horst

This is how to deal with visualization urges: 6 tips for designing visualizations

Improvements in data quality drive process optimization, creating an upward spiral.

Data in order? Improve your processes too

Data op orde

Data is increasingly put in order as a result of attention to data management. It is forgotten that this leads to improvements in the primary and supporting processes, resulting in fewer errors and higher effectiveness. These process improvements, in turn, ensure that the data quality increases, resulting in a self-reinforcing effect. So after putting the data in order, don't forget to improve the processes. 

I meet more and more organizations that are serious about data management. This is to comply with the AVG (General Data Protection Regulation) but also because they can realize their ambitions such as information-driven work on the basis of data. When complying with the AVG, the focus is on what data may and must be retained and who has access to this data. With information-driven work, it is mainly the search for data on which control can take place. 

As a result, data has increasingly been put in order. In other words, an inventory is made of the data present within the organization, because there is often no overview of it, and then it is placed in a suitable location with appropriate authorizations. But all too often I see that it is then forgotten to make better use of this data in the primary and supported processes.  An open goal is missed. 

An example of this in the primary process is putting client files in order in a healthcare institution. The added value of this becomes much greater if one then looks at how this data can be used effectively and efficiently by employees. What information do they need in order to best help the client in specific cases? How can we ensure that precisely this information is easily and quickly extracted from the file? 

Another example is the deployment of steering information. Based on this data, the quality of the capacity management of employees in support processes can be examined, or where in administrative processes the greatest delays occur. This often turns out to be low-hanging fruit because it takes little time to identify improvements. 

and if process improvements are achieved in this way, this also leads retroactively to further strengthening data quality. After all, when processes have fewer errors and are more effective, data quality also increases. This creates a self-reinforcing effect. 

If you want to know where data-based process improvement can be achieved for your organization, please contact me.

Data visualization drives data-driven decisions: Insights, efficiency, informed choices.

Storytelling with datavisualization

Storytelling met datavisualisatie

Governments want to work in a data-driven way, using data to design policy and implementation more efficiently. This starts with making the data accessible, offering insight into the data and thereby telling the story. Data visualization and dashboards offer a low-threshold method for this, and directly help to work and steer in a data-driven manner.

Ethical oversight crucial in AI integration. Data Ethics Officer ensures ethical standards

Does every business needs a Ethics Officer?

AI-Ethics: Does every organization need an Ethics Officer?

Steven Spielberg's "Minority Report" outlines a dystopian vision of the future in which a few individuals can predict crimes before they happen. Interesting, but what if these predictions are wrong, as happened in the film on occasion? These kinds of questions are now commonplace outside of Hollywood. Thanks to advanced computer technology, increasingly sophisticated algorithms are being developed. The past decade has taught us that risk in its application has not always received the attention it needed in hindsight. For example, "Follow The Money" recently revealed that until 2015, the Dutch police used an algorithm that estimated the risk of criminal behavior based in part on ethnicity. This could lead to preventive and repressive measures such as a person-centered approach. A person-centered approach can mean faster police action, such as preventive searches, confiscating money and valuables, and even providing mandatory care. In addition, families can be intervened with to prevent other family members, such as siblings, from taking the wrong path as well. Organizations have also used predictive algorithms in other public initiatives, such as SyRI and the benefits affair, sometimes with far-reaching consequences. 
With the rise of artificial intelligence, such as Chat-GPT, we can analyze huge data sets and identify patterns. However, the increasing power of AI also makes it challenging to assess the ethical aspects of an algorithm. 
While predictive systems can be valuable, proper oversight is essential to prevent (unintended) abuses. Given the complexity of these systems, I emphasize the importance of data ethics within public organizations, especially when combined with AI. We call this AI-ethics. This can start with the appointment of a data ethics specialist, with which the organization formally invests this responsibility. In this article, therefore, I discuss the role of a Data Ethics Officer.

Highberg case study: Municipality of Rotterdam responsible use algorithms and AI systems

Responsible use of algorithms and AI systems

Highberg ondersteunt de gemeente Rotterdam bij verantwoord gebruik van algoritmen en AI systemen

Highberg digital transformation has assisted the municipality of Rotterdam in the responsible use of algorithms and transparency about these algorithms. In this project our experts managed the algorithm registry and executed the processes in the algorithm governance. During the project our experts provided concrete improvement suggestions and implemented these improvements in practice.

Highberg supports the municipality of Rotterdam in the responsible use of algorithms and AI systems

Media and entertainment

Which ones are you already using?

Key AI use cases in newspaper publishing

the-most-important-ai-use-cases-from-newspaper-publishers-which-ones-are-you-already-using

Artificial intelligence (AI) will be one of the strategic cornerstones of economic success for newspaper publishers in the coming years. This is the opinion of 77% of publishers surveyed in a global study conducted by Highberg (formerly known as Schickler) together with WAN-IFRA on the use of AI at newspaper publishers. But how do you actually use AI properly and which areas should you focus on? 
With many AI use cases already successfully in use at global industry pioneers, the answer to this is now well known. We have summarized the results of our global study on the most important AI use cases in the areas of reader market, content production and content playout for you below.

Ole Fehling

The most important AI use cases from newspaper publishers – which ones are you already using?

The foundation of your digital future

Data Lakehouse

Data-Lakehouse-The-foundation-of-your-digital-future

The importance of data for a digital strategy is undisputed. The problem: The data is often not available, not processable or not complete. The cause usually lies in the traditional IT infrastructure. The modern solution is a data lakehouse, as an evolution of the data warehouse.

Rolf-Dieter Lafrenz

Data Lakehouse - The foundation of your digital future

Optimization of strategy, culture, organization and technology

Data strategy in media

transformation-into-a-data-driven-media-company

Transformation to a data-driven media company: Why the right strategy, culture, and organizational form are more important than the technological foundation.

Transformation into a data-driven media company

A Highberg customer story: Deciding with the help of data analysis at SZW

Deciding with help of data analysis at SZW

Laptop data lukas blazek mc S Dtb WXUZU unsplash

The Dutch Social Affairs and Employment Inspectorate (Inspectie SZW) was created in 2012 by the merger of the Dutch Labor Inspectorate, the Work and Income Inspectorate and the Social Intelligence and Investigation Service. It is the supervisor in the policy areas of Working Conditions, Labor Market, Labor Relations and Social Security in the Netherlands. As a supervisor, Inspectorate SZW conducts inspections at businesses, such as restaurants and construction sites. In addition, this government body also conducts investigation and enforcement, for example in the area of benefit fraud. This work is carried out on the basis of risk and environmental analyses.

Christian Verhagen

Deciding with the help of data analysis at SZW

Microsoft 365 transition: Challenges and tips for data migration

Microsoft 365 & the data migration challenge

the-modern-microsoft-365-workspace-and-the-data-migration-challenge

We are witnessing a rapid replacement of old 'thin client solutions' with modern Microsoft 365 workspaces. No more RDS or VDI, but fat clients with modern cloud-based Office solutions. Equally secure and manageable as RDS, but faster, user-friendly, and more mobile. Experience has shown that there are a few points of attention. Today: the challenges of migrating data to the new solution (SharePoint and OneDrive).

Cleo van Engelen

The modern Microsoft 365 workspace and the data migration challenge

Prepare IoT projects thoroughly: Define goals, involve stakeholders, manage risks.

The Internet of Things and what to expect

More than ever, technological developments are bringing with them a variety of possibilities. Because more is possible, organizations also need to know more about it to determine what to do or not to do. In most cases, developments can work for you, but thorough preparation before starting projects is recommended to avoid disappointment. So what should you look for when starting a project with an Internet of Things (IoT) component? 

An organization is caught off guard when decisions must be made under apparent time pressure to "do something" with available technologies. Think blockchain, machine learning, 3d printing, drones, etc. IoT is not new, but the applications from organizations are often still in their infancy. In the case of IoT, the question is, "You want a 'smart' organization too, right?" You hear that amazing things are possible. The application goes beyond sensors, the price is more attractive than before and it can all be done wirelessly. So why not just try it out in a project? 

Unfortunately, practice is more recalcitrant, according to a recently published Cisco survey of IoT projects. 1845 ICT decision makers were surveyed about their experiences with IoT. They indicated that nearly 75 percent of projects do not lead to the desired results. This starts as early as the Proof of Concept, where expectations are not in line with the result. Top management generally does not know what the technology entails and what the possibilities are, strategy is not connected to operations, business and ICT do not work together and expertise is insufficient. 

Launching IoT projects should be in line with the organization's vision and strategy. But what are the considerations and tools to make a good decision about this? In my opinion, preparation is essential in this regard. Answering important questions in the preparation gives a grip on the result. Examples of such questions, observed from practice, are: 

- What exactly is IoT? 
- What business objective is IoT actually being used for? 
- What is the added value of the project for the organization? 
- How is IoT related to other developments inside and outside the organization? 
- Is there sufficient knowledge of and connection between strategic, tactical and operational layers within the organization regarding IoT? 
- What should be the role of management? 
- What should the collaboration between business and ICT look like? 
- For what purposes might malicious parties be able to misuse an IoT application? 
- How is privacy regulated? Read more about this in VKA consultant Bas de Groot's blog on how to deal with the approaching European (e-)privacy regulation. 
- How stable are networks to choose? 
- How is the integration with IoT cloud platforms? 

Involve the right stakeholders in these questions, from strategy to operations. And above all, let them be "live" questions, so that they are asked throughout the project. This creates a better grip on whether it is in line with expectations. For example, start with workshops weighing opportunities and risks. Then prioritize the most important considerations. This

Navigating the complexities of upscaling data & analytics in civil society organizations

Data & analytics: Beyond the pilot phase

Data-analytics-Beyond-the-pilot-phase

On February 3rd 2023, we organized a Round Table on Data &amp; Analytics for civil society organisations. Large and small municipalities, provinces, ministries and the House of Representatives joined the virtual table. The central question that was at the table: How do we get from pilot to upscaling? 
Many organizations have set up innovation teams or data labs. After initial successes, it often proves to be a challenge to ensure that data applications are well embedded in the organization, are scalable and are developed further in order to actually realize the added value of the new data application for the business. In this roundtable, we presented four dilemmas and discussed success factors for technical and organizational upscaling. With some additions from our side, we give a view of the afternoon in this insight.

Sander Spanjaard

Footer Navigation

Header Navigation

Responsible AI

Digital Law

Data Management & Governance

Sabine Steenwinkel den Daas

Senior Consultant Digital Law & Ethics

Sabine has ten years of experience in developing, implementing, validating, and assessing data and AI applications. She possesses a strong analytical and theoretical foundation to understand algorithms and AI deeply. With her background in data science, entrepreneurship, numerous IT project assignments, and certifications in privacy (CIPM) and data management (CDMP), Sabine comprehends the necessary prerequisites and control measures for successful AI implementations in great detail. She is also the substantive founder of the Highberg (c) assessment and standards framework for algorithms and has tested multiple AI systems in practice. Sabine is aware of the latest developments, such as the AI Act, and knows the ISO/IEC 42001 in detail. In her daily work, Sabine provides concrete advice on how to stay in control of data and algorithms in a responsible way, and she also enjoys sharing her knowledge through training sessions and webinars. Want to know more? Connect with Sabine on <a href="https://www.linkedin.com/in/sabinedendaas/" target="_blank" rel="noreferrer noopener">LinkedIn</a>.

Sabine Steenwinkel-den Daas

Senior Consultant Digital Law & Ethics

Capabilties

Gerelateerde Insights