[et_pb_section fb_built=\"1\u2033 admin_label=\"section\" _builder_version=\"4.23\u2033 background_color=\"#000000\u2033 custom_padding=\"0px||||false|false\" global_colors_info=\"{}\"][et_pb_row _builder_version=\"4.27.4\u2033 _module_preset=\"default\" width=\"100%\" custom_padding=\"||0px||false|false\" global_colors_info=\"{}\"][et_pb_column type=\"4_4\u2033 _builder_version=\"4.23\u2033 _module_preset=\"default\" global_colors_info=\"{}\"][et_pb_text _builder_version=\"4.27.4\u2033 _module_preset=\"default\" global_colors_info=\"{}\"]<\/p>\n
Artificial intelligence (AI) is high on the agenda for many organizations, including in socially important sectors such as healthcare, education and government. Rightly so: the challenges are enormous, but fortunately so are the technical possibilities. AI can help doctors treat patients better and more personally, support schools in personalized learning processes and help governments make services more efficient and effective.<\/span>\u00a0<\/span><\/p>\n But the social impact is great. These sectors in particular involve vulnerable groups and important public values. An incorrect prediction can lead to unequal treatment, reputational damage or loss of trust among citizens and regulators. This is why the EU AI Act labels many of these applications as high risk.<\/span>\u00a0<\/span><\/p>\n That sounds harsh, but it is certainly not a blockade to using AI. On the contrary, those who invest in honesty, transparency and sustainable quality of AI now are building not only compliance, but also trust and innovativeness.<\/span>\u00a0<\/span><\/p>\n [\/et_pb_text][et_pb_text _builder_version=\"4.27.4\u2033 _module_preset=\"default\" global_colors_info=\"{}\"]<\/p>\n [\/et_pb_text][et_pb_text _builder_version=\"4.27.4\u2033 _module_preset=\"default\" global_colors_info=\"{}\"]<\/p>\n Responsible AI use requires more than just technical solutions. It is about finding the right balance between innovation and diligence. The basis for this lies in three pillars: honesty and ethics, safety and transparency, and sustainable quality.\u00a0<\/span><\/span>\u00a0<\/span><\/p>\n [\/et_pb_text][et_pb_text _builder_version=\"4.27.4\u2033 _module_preset=\"default\" custom_padding=\"||0px|||\" global_colors_info=\"{}\"]<\/p>\n 1. Honesty and ethics<\/span><\/span>\u00a0<\/span><\/strong><\/p>\n AI learns from historical data. And that data often contains existing inequality. The risk is that AI reinforces those inequalities. Consider a hospital that is more likely to label patients from certain neighborhoods as \"no-shows,\" or an educational institution that underestimates students from immigrant backgrounds.<\/span>\u00a0<\/span><\/p>\n The good news: honesty is measurable and manageable! Tools such as\u00a0<\/span>Fairlearn<\/span><\/b> show whether a model provides equal opportunities for different groups. They signal whether error margins are skewed and offer ways to correct that. This prevents a clever prediction from becoming an unfair practice.<\/span><\/p>\n [\/et_pb_text][et_pb_blurb _builder_version=\"4.27.4\u2033 _module_preset=\"default\" global_colors_info=\"{}\"][\/et_pb_blurb][\/et_pb_column][\/et_pb_row][et_pb_row _builder_version=\"4.27.4\u2033 _module_preset=\"default\" width=\"100%\" min_height=\"758px\" custom_padding=\"||||false|false\" global_colors_info=\"{}\"][et_pb_column type=\"4_4\u2033 _builder_version=\"4.23\u2033 _module_preset=\"default\" global_colors_info=\"{}\"][et_pb_testimonial quote_icon_background_color=\"#49b69c\" _builder_version=\"4.27.4\u2033 _module_preset=\"default\" body_text_color=\"#FFFFFF\" author_font=\"PP Neue Machina Ink Ultrabold|700|||||||\" background_color=\"RGBA(255,255,255,0)\" border_width_all=\"1px\" border_color_all=\"#47B2A8\u2033 border_color_all_portrait=\"RGBA(255,255,255,0)\" box_shadow_style=\"preset1\u2033 box_shadow_color=\"#49b69c\" global_colors_info=\"{}\"]<\/p>\n Also in our projects we apply\u00a0<\/span><\/span>fairness<\/span>-techniques in areas where the social impact is high. Organizations that take this seriously not only comply with regulations, but also gain support from citizens, clients or employees by communicating openly and transparently.<\/span><\/span>\u00a0<\/span><\/p>\n [\/et_pb_testimonial][et_pb_text _builder_version=\"4.27.4\u2033 _module_preset=\"default\" custom_padding=\"||0px|||\" global_colors_info=\"{}\"]<\/p>\n 2. Security and transparency<\/span><\/span><\/strong><\/p>\n An AI model that supports a decision without explanation will not work in industries where trust is crucial. This is also true of AI models that involve human intervention. Physicians, educators and policy makers need to understand why a model gives a particular outcome in order to act on it responsibly.<\/span>\u00a0<\/span><\/p>\n That's why there are techniques that make AI transparent.\u00a0<\/span>SHAP<\/span><\/b>\u00a0and\u00a0<\/span>LIME<\/span><\/b>\u00a0are examples of methods that show what factors influence a decision. With these, a doctor can substantiate why a patient is at extra risk, or a teacher can explain why a student receives extra support. Transparency makes AI not only explainable, but also governable. In addition, understanding the relationship between an outcome and certain factors provides all kinds of new insights that would never have been found without these kinds of complex algorithms.\u00a0<\/span>\u00a0<\/span><\/p>\n Especially with language models, which are gaining tremendous interest thanks to ChatGPT and Copilot, it is important to make a conscious decision to choose the right type of AI model.\u00a0<\/span>Encoder models<\/span><\/b>\u00a0(such as BERT for the Dutch language) are strong in analyzing and understanding text and therefore often more explainable.\u00a0<\/span>Decoder Models<\/span><\/b>\u00a0(such as GPT) can generate creative responses and simulate a conversation, but their reasoning is very difficult to fathom. For industries where honesty and explainability outweigh creativity, an encoder model is often the wiser choice.<\/span>\u00a0<\/span><\/p>\n [\/et_pb_text][et_pb_text _builder_version=\"4.27.4\u2033 _module_preset=\"default\" custom_padding=\"0px||0px|||\" global_colors_info=\"{}\"]<\/p>\n 3. Sustainable quality<\/span><\/span>\u00a0<\/span><\/strong><\/p>\n A model that performs well today may become obsolete tomorrow. Data changes, context shifts. This phenomenon, model drift, can lead to wrong predictions if it is not noticed.<\/span>\u00a0<\/span><\/p>\n Monitoring is therefore essential. Tools such as\u00a0<\/span>Popmon<\/span><\/b>\u00a0(Population Monitoring) continuously compare new data with reference data and signal when values begin to shift. This can result in temporary disabling or retraining of the model. There are tempting techniques for automatic retraining, but without human supervision, new bias can then be embedded, so be careful with that.<\/span>\u00a0<\/span><\/p>\n For our customers, we build monitoring processes that combine technology and governance. This way we ensure that models are reliable not only today, but also in the future.<\/span>\u00a0<\/span><\/p>\n [\/et_pb_text][et_pb_text _builder_version=\"4.27.4\u2033 _module_preset=\"default\" global_colors_info=\"{}\"]<\/p>\n [\/et_pb_text][et_pb_text _builder_version=\"4.27.4\u2033 _module_preset=\"default\" custom_padding=\"0px||0px|||\" global_colors_info=\"{}\"]<\/p>\n AI has the potential to make healthcare smarter, education fairer and government more efficient. But that can only happen if we use AI responsibly. With the right approach, the EU AI Act will not become a barrier, but rather a catalyst for better, more reliable technology.<\/span>\u00a0<\/span><\/p>\n <\/span><\/p>\n [\/et_pb_text][et_pb_testimonial quote_icon_background_color=\"#49b69c\" _builder_version=\"4.27.4\u2033 _module_preset=\"default\" body_text_color=\"#FFFFFF\" author_font=\"PP Neue Machina Ink Ultrabold|700|||||||\" background_color=\"RGBA(255,255,255,0)\" border_width_all=\"1px\" border_color_all=\"#47B2A8\u2033 border_color_all_portrait=\"RGBA(255,255,255,0)\" box_shadow_style=\"preset1\u2033 box_shadow_color=\"#49b69c\" global_colors_info=\"{}\"]<\/p>\n For decision-makers and consultants, the core message is clear: responsible AI use is a strategic prerequisite and, in many cases, quite achievable as long as you work at it with attention.<\/span><\/span><\/p>\n [\/et_pb_testimonial][et_pb_text _builder_version=\"4.27.4\u2033 _module_preset=\"default\" custom_padding=\"0px||0px|||\" global_colors_info=\"{}\"]<\/p>\n AI has the potential to make healthcare smarter, education fairer and government more efficient. But that can only happen if we use AI responsibly. With the right approach, the EU AI Act will not become a barrier, but rather a catalyst for better, more reliable technology.<\/span>\u00a0<\/span><\/p>\n From our experience, we know how important it is to combine technical innovation with careful governance, clear frameworks and support within the organization. We help organizations develop and apply AI solutions that are not only effective, but also fair, transparent and future-proof. This is how we build a future together in which AI truly adds value to people and society.<\/span>\u00a0<\/span><\/p>\n <\/span><\/p>\n [\/et_pb_text][\/et_pb_column][\/et_pb_row][\/et_pb_section].<\/p>","protected":false},"excerpt":{"rendered":" Find out in this article how to deploy AI honestly, transparently and responsibly without inhibiting innovation.<\/p>","protected":false},"author":3,"featured_media":10887,"comment_status":"closed","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"_et_pb_use_builder":"on","_et_pb_old_content":"","_et_gb_content_width":"","footnotes":""},"categories":[82,51,24,76,55],"tags":[],"class_list":["post-10846","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-article","category-artikel","category-blog","category-data-science-ai","category-data-science-ai-nl"],"_links":{"self":[{"href":"https:\/\/cmotions.com\/en\/wp-json\/wp\/v2\/posts\/10846","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/cmotions.com\/en\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/cmotions.com\/en\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/cmotions.com\/en\/wp-json\/wp\/v2\/users\/3"}],"replies":[{"embeddable":true,"href":"https:\/\/cmotions.com\/en\/wp-json\/wp\/v2\/comments?post=10846"}],"version-history":[{"count":28,"href":"https:\/\/cmotions.com\/en\/wp-json\/wp\/v2\/posts\/10846\/revisions"}],"predecessor-version":[{"id":10899,"href":"https:\/\/cmotions.com\/en\/wp-json\/wp\/v2\/posts\/10846\/revisions\/10899"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/cmotions.com\/en\/wp-json\/wp\/v2\/media\/10887"}],"wp:attachment":[{"href":"https:\/\/cmotions.com\/en\/wp-json\/wp\/v2\/media?parent=10846"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/cmotions.com\/en\/wp-json\/wp\/v2\/categories?post=10846"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/cmotions.com\/en\/wp-json\/wp\/v2\/tags?post=10846"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}Three pillars of responsible AI<\/h2>\n
Building responsible AI together<\/span><\/span><\/h2>\n