Beyond the Algorithm: Why Human Linguistic Expertise is Essential for True Understanding and Impact

Verbal reasoning tests our ability to understand, evaluate, and derive logical conclusions from written language. Which means that the way something is written or said, profoundly influences our understanding, decision making and possible actions. To assess verbal reasoning skills, tests are usually drafted in the mother tongue of the test-sitter (save for mobile citizens) and the multilingual aspect is often ignored.

But when in a multilingual context, it makes sense to consider the effect of copy on the verbal reasoning of the final consumer of such information. A standardised, flattened, morphologically simplified, often incongruent or recursive language, offered by current machine translation tools and content generators, may cause cognitive dissonances  that impact learning, decision-making, and reasoning.

The painstaking work of linguists (and not process or system integration experts) is what makes the difference between proper and poor understanding, correct and wrong decisions, success or mistake, high or no risk. It is an aspect which often goes under among the enthusiasm of those that are not trained in the multimodality of simple words (aka, engineers). But it is what makes a message effective.

Following, a list of variables and their impact on verbal reasoning in a mono- and in a multilingual setting.

1. Textual Clarity

• Impact on reasoning:
  Unclear texts cause confusion, misunderstanding, or misinterpretation. Cognitive resources are spent decoding rather than analysing meaning. The latter often happens in AI generated or text translated with machine translation.

• Example:
  Ambiguous phrasing or overly complex syntax can obscure logical relationships, making it hard to grasp intended arguments.

• In multilingual contexts:
  Clarity diminishes further due to inadequate translation, inappropriate equivalences, or unclear syntax (i.e. the 1:1 correspondence in Machine Translation). To this point, it is important to note that translators are trained to convey the meaning - not necessarily the words - inferred from the source text, that is, the message behind a text, in consideration of its intent and its purpose. Often,  a 1:1 correspondence between words does not cut it. But this is what MT and automatic QA tools do. So beware of what you ask.

2. Coherence and Cohesion

• Impact on reasoning:
  Coherent texts provide logical connections and smooth, effortless cognitive processing. Lack of coherence interrupts reasoning flow and diminishes retention. 

• Example:
  Poorly linked ideas (e.g., weak transitions, inconsistent argumentation) disrupt the logical narrative, making it recursive and self-referencing without really adding information (we).

• In multilingual contexts:
  Some languages follow different fragmentation and sequencing rules that, if not adopted, significantly impair logic and understanding (think about the focus, intent and meaning of AI generated texts or machine translated texts). Although more "attentive" to the context, automation tools still focus on siloed information, segmented at sentence level and often extrapolated from its context, with translation often focussed on deltas/changes and other very relevant information for cohesiveness is simply obscured (it costs less, right?).

3. Terminology and Lexical Precision

• Impact on reasoning:
  Precise terminology anchors concepts clearly in readers’ minds. Inaccurate terms create misunderstanding or ambiguity. Terminological work is central for disambiguation and targeted communication.

• Example:
  Technical or domain-specific terms incorrectly used can distort comprehension (e.g., legal terms mistranslated in cross-jurisdictional contracts, where translated version have the same validity as the original copies).

• In multilingual contexts:
  Misalignment of terminology across languages (particularly in specialized fields) can create severe cognitive dissonances, resulting in faulty reasoning, wrong decision, insufficient reach, with a detrimental effect also on data used further down the line to train AI tools.

4. Phrase Structure and Syntax

• Impact on reasoning:
  Complexity of syntactical structures (long, convoluted sentences) significantly increase cognitive load, reducing working memory available for higher-order reasoning (waiting for a verb in German can be exhausting).

• Example:
  Sentences that exceed cognitive capacity make readers forget earlier parts by the time they reach the end, obscuring overall meaning. But keep in mind that in some cultures, longer sentences are not only acceptable, but required in certain types of communication.

• In multilingual contexts:
  Translations that preserve source-language structures rather than adapting to target-language norms compound difficulty, leading to decreased comprehension and generating estrangement and distance, which ultimately diminishes the end-user engagement.

5. Style, Register, and Tone

• Impact on reasoning:
  A style suited to audience expectations enhances reasoning ability. Style mismatch can reduce cognitive engagement and credibility of messages.

• Example:
  An overly informal style in a serious context can reduce reader trust or attention, thereby diminishing the impact of logical arguments. And the degree of acceptance for informality varies from country to country and as such needs to be adapted.

• In multilingual contexts:
  Style variations—if not adapted culturally and linguistically—may produce confusion, weakening the reader's engagement, trust and cognitive absorption. Puns, idioms and metaphors may work in a language but not in another, leading to miscommunication. Obviously, Machine Translation translates them verbatim. 

6. Accessibility and Readability

• Impact on reasoning:
  Easily accessible texts maximize cognitive efficiency and enable deeper comprehension. Difficult texts create barriers, reducing cognitive bandwidth for reasoning. So always consider the purpose of your communication, and make it accessible - in linguistic/cognitive terms not as technical data availability/accessibility, which is a corollary.

• Example:
  Readability improvements (shorter paragraphs, bulleted lists, limited use of acronyms) directly correlate with better retention and understanding (but not all cultures welcome such structured approach).

• In multilingual contexts:
  Readability problems are often amplified due to literal translations, culturally inappropriate formats, or unknown acronyms significantly obstruct learning and comprehension.  

7. Cultural Appropriateness

• Impact on reasoning:
  Cultural nuances affect interpretation. Texts that are culturally incongruent can disrupt meaning-making processes, causing misunderstandings and wrong decision taking.

• Example:
  Metaphors or idioms translated literally cause confusion or disrupt the narrative logic.

• In multilingual contexts:
  Without careful cultural adaptation, reasoning is impeded because readers struggle to find logical consistency or relevance. And what is not relevant, is not remembered. Further, the use of hyperboles or understatement differs from culture to culture and if not "translated" can lead to misunderstandings with dramatic consequences.

The multilingual perspective

To sum it up, the end reader in a different language might experience the following:

• Cognitive dissonance due to translation errors:
  Readers face internal conflict, reconciling contradictory or poorly explained logical steps, impairing their reasoning processes and understanding.

• Interference between languages (linguistic transfer):
  Readers processing multilingual materials often experience cognitive overload due to subconscious comparisons or interference, causing weakened attention to logical details.

• Loss of trust in the message and lack of persuasion power:
  Multilingual dissonance leads to reduced trustworthiness of arguments, undermining impact and persuasive power.

What does this mean in practice?

• Cognitive load increases, attention span shortens.

• Learning is shallow, retention low.

• Readers’ reasoning capacities (ability to infer, generalize, analyse, synthesise) are limited due to effort spent decoding the surface-level meaning.

• Misunderstandings result in decision-making errors or misguided conclusions, sometimes with dramatic consequences.

• Text’s intended purpose (educational, persuasive, procedural) is not achieved, resulting in wasted effort and potentially serious losses.

So, what multicultural best practices can you adopt?

1. Simplify syntax and sentence structure to ease cognitive load, but consider the cultural element, addressees and preferences in each cultural space.

2. Clarify lexical choice to minimize ambiguity and misunderstanding. Consider terminology as if it was a branding guide. Sometimes it is good to offer a glossary to avoid misunderstandings that could bear consequences (like the Definitions section in contractual documents).

3. Ensure coherence and logical progression between ideas, making explicit connections visible. But again, do not forget the intercultural element: some cultures love to read between the lines.

4. Adapt cultural and stylistic elements, adapt them to the audience’s expectations, in  the relevant country.

5. Verify terminology consistency to maintain logical integrity in multilingual environments.

6. Evaluate readability metrics to improve accessibility, as relevant in the specific country (i.e. an over-simplified text may praised in English, but considered too simplistic in another language).

7. Incorporate quality linguistic reviews and cognitive testing of multilingual drafts to identify and address cognitive dissonance early.

8. Avoid using poor data for AI tools training.

Conclusion

The linguistic and textual characteristics of copy significantly affect our verbal reasoning capabilities. Clear, coherent, culturally-adapted, and cognitively optimized language (the job of a linguist) facilitates deeper (and mutual) understanding, logical reasoning, effective learning and sound decision-making.
Conversely, perfunctory or inadequately localised texts weaken our reasoning processes, limit understanding, introduce an element of risk and negatively influence both decision-making outcomes and relationships. This represents an impoverishment that extends well beyond immediate business concerns, affecting our broader capacity for reasoning and our most innate ability to express ourselves as human beings.

If language is what allows us, humans, to create constructive relationships and progress, even build our reality, why opt for a  multilingual approach that fattens, standardises and transforms our hard-thought intent into a probabilistic outcome?

Credits: picture of brain generated with AI