In the long tradition of claims in expertise that are not, we have the latest from Safety as non-experts in Linguistics (https://safetyrisk.net/more-of-safety-talking-to-safety-about-safety/). Apparently, counting repetition of words as data is ‘language intelligence’!
Of course, it isn’t, but why would that bother Safety when it has people to blame and victims to harm.
Next is what this article calls ‘The Paper Problem’ and, one might think this might be about the excess of paperwork associated with safety? Nup, the problem is about the quality of content in conversations. Again, the problem is cultural.
Apparently, the way to fix these culture problems is to:
‘objectively measuring and optimizing these conversations at scale has proved an elusive challenge’
Then it jumps to this absurd proposition:
‘Recent technological innovations in artificial intelligence (AI), machine learning and natural language processing (NLP) are transforming these daily planning conversations from paper checkboxes into catalysts for operational excellence (Pettinger & Nelson, 2024).’
Of course, if one was thinking about Linguistic Intelligence and Semantics (not mentioned) one would expect the bibliography to be littered with research anchored to Lacan, Kristeva, Jakobson or Halliday. One might expect the fundamentals of understanding grammar, discourse analysis, Sociolinguistics, Phonology, Psycholinguistics, Morphosyntax, Functional Grammar, Language acquisition, comprehension and reading formation to be included in discussion, of course not.
All of the essentials required in studying English at University as a pre-requisite for Linguistics are absent from this article. Everything I learned in my English major in my first degree to become an English teacher is absent from this article. Indeed, the writers have no expertise in Linguistics, Semiotics or Semantics. Ah, that’s the safety way.
If one is interested in the fundamentals in language, the last place to seek any intelligence in Linguistics is Safety.
If you are interested in developing intelligence in Language, perhaps start here:
- Jakobson & Halle (1971) The Fundamentals of Language (https://monoskop.org/images/c/cc/Jakobson_Roman_Halle_Morris_Fundamentals_of_Language_2nd_ed_1971.pdf)
- Fasold &Conner-Linton (2006) An Introduction to language and Linguistics (https://uogbooks.wordpress.com/wp-content/uploads/2014/10/fasold_r._connor-linton_j._an_introduction_to_lz-lib.org_.pdf)
- Fromkin, Rodman and Hyams., (2014). An Introduction to Language. (https://ukhtt3nee.wordpress.com/wp-content/uploads/2019/04/an_introduction_to_language.pdf )
What is typical of Safety as ‘expert in everything’ is that all of what is contained in these texts is missing from this so-called article on so-called ‘Language Intelligence’. Indeed, this article has nothing to do with language intelligence.
The first and most critical danger of this article is the nonsense idea that one can learn about language through AI, NLP and machine non-learning. The repetition of algorithms, viewing language as data and using computers to scan text teaches nothing about the basics of Linguistics and Language. AI has no knowledge of any emotion anchored to text, it cannot feel the meaning of a language, interpret context and has no idea what emotions are conveyed in Poetics (all that is non-measurable).
The next nonsense in this article is about ‘scoring’ conversations. There is no way of giving a score to a conversation. All speech, text and language are qualitative not quantitative. All language is interpreted and translated, meaning in text is understood by con-text.
No AI or computer can ‘understand’ the nuances of language in context because a computer cannot ‘live’ in context not understand culture. For example. How can AI understand the meaning of First Nations Dreamtime? What can AI know about dreaming? When AI is switched off, it has no dreams. There is no embodiment in computing because computers have no bodies. How can a computer understand and empathise with the emotions expressed in text? How can a computer understand how a risk is felt?
Computers cannot embody e-motion so, they cannot learn.
The next bit of nonsense in this article is the idea of ‘assessing conversation quality by scale’. The idea that language translation is transcription is simply silly. Converting sound to text is not translation. Of course, this article doesn’t discuss the problem of machines ‘interpreting’ text without an embodied understanding of culture. Unless a computer is programmed a certain way, it can never ‘understand’ the nuances of Australian idioms, slang or cultural language.
And of course, machines can’t listen.
When we listen to language we bring together years of experiences learned through culture to understand when ‘no’ means ‘no’ and when ‘no’ means ‘yes’. If you are interested in interpretation, perhaps have a look at how the word ‘f$ck’ can be interpreted 26 ways (https://reallifeglobal.com/how-use-word-fuck/). On most building sites I’ve been on that word is: a conjunction, preposition, adjective, noun, verb, participle and prefix.
One of the best ways to ensure that people don’t speak up is to ‘score’ conversations. When people learn that their words will be scored and put into a computer and then interpreted by people with power, all conversation ceases. For the same reason, this is why workers with personal concerns will never speak to a Zero Harm Advisor or a supervisor in a zero organisation. The last place to find empathy, care, understanding, support or counselling skills is in Safety=Zero.
The more you progress through this article and its lack of expertise, the more you realise the many mythologies it creates. For example:
‘By collecting, transcribing, structuring and analyzing conversations, the company’s leaders can uncover hidden opportunities for improvement, foster authentic conversations, and ultimately build a stronger, safer and more prosperous workplace’.
Of course, this is just nonsense projection without evidence. At a cultural level any ‘collecting, transcribing, structuring and analyzing conversations’ will shut conversations down.
Then this:
‘AI technology and data science have transformed the way company leaders view daily planning conversations, revealing their significance in shaping organizational culture and safety performance. “This innovative approach is not just about technology; it’s about the power of understanding and harnessing the insights that lie within our daily interactions’
So, what we see is the accumulation of data to control employees. In this there is no discussion of ethics, culture, power, politics or legal implications. It’s whole argument is founded on Linguistic gnorance.
All this article will foster is continued secrecy, lack of engagement and cultural ignorance. Any company that seeks to implement the recommendations of this article will go backwards in social cohesion, trust and relational intelligence.
When the article states:
‘By decoding the daily conversations using innovative language AI and machine learning, the safety professional can transition from an overreliance on enforcement-based safety driven by lagging metrics to an approach centered on fostering respect, gratitude, trust and collective hazard ownership through the creation of environments with a higher probability for psychologically safe conversations.’
What it means is the opposite.
Monitoring, analysing and scoring conversations is a recipe for a lack of trust, a lack of respect and the suppression of confidentiality essential for psychosocial safety!
As usual, this article is more Safety projection, assertions and attribution for its opposite. This is what typically happens when Safety assumes expertise it doesn’t have.
The best way to improve engagement, conversations and trust for psychosocial safety, is to focus on developing a culture of empathy, mutuality, listening and dialogue. This is because:
- Language is not data to be manipulated for some loaded safety outcome.
- Any analysis of language requires extensive Linguistic expertise that is culturally human and NOT based in computing technology.
- Using AI and non-machine learning disrupts relationships, social trust and engagement.
- Computers don’t have social relationships with persons or their emotional being.
- AI is not a social tool.
- There is no such thing as ‘computational linguistics’.
- Computers don’t learn, acquire or speak language like humans. They are programmed and don’t learn, nor do they acquire language by experience as humans do.
- AI and non-machine learning don’t have a ‘perception’ or ‘unconscious’ with which to understand persons or what they say.
The best conversations in safety are:
- Informal and not counted or analysed.
- Based on mutuality and sensitivity in understanding, based upon nuances learned through relationships.
- When those in leadership understand that humility, social meaning, trust and culture are the beginning of connection and dialogue.
- When leaders suspend their agenda (safety) and engage others in relationships of trust where power is given to ‘the other’.
- About person to person skills not anything to do with computers.
- When humans get away from computers and walk and listen without an agenda.
This article has no expertise in any of this. Neither, is there anything in the safety curriculum globally that supports the building of relationships, enhancing personhood, understanding language and culture or the creation of Linguistic Intelligence.
In SPoR, we help teach all these essentials in Language and Linguistics through iCue Engagement skills (https://safetyrisk.net/icue-as-visual-verbal-risk-assessment-a-video/ ). This is why we don’t analyse conversations but ‘map’ them and, hand power to the participant to control the conversation (https://safetyrisk.net/risk-icue-video/). The learning associated with developing Linguistics skills on a foundation of Social Psychology: builds trust, relationships and engagement in conversations that matter.
In SPoR, we also teach the basics in Linguistics (https://cllr.com.au/product/linguistics-flyer-unit-21/) from a basis of expertise in English, Language, Para-Linguistics and Semiotics (https://safetyrisk.net/culture-cloud-tour-part-9-language-linguistics-and-paralinguistics/). All of this learning is positive, practical and works (https://www.humandymensions.com/product/it-work-a-new-approach-to-risk-and-safety-book-for-free-download/).
Dr Rob Long
Latest posts by Dr Rob Long (see all)
- Critical Thinking Needed More Than Ever in Safety - January 21, 2025
- Selling Safety and the Safety of Selling Safety Culture - January 19, 2025
- Hearing Voices - January 18, 2025
- You Can’t ‘Bust’ Myths - January 18, 2025
- There is No Such Thing as an HRO - January 18, 2025
Brent Charlton says
So much in human communication relies on body language, the eyes, facial expression, tone, etc. Computers can’t pick up all that nuance.
Rob Long says
Yes, and that’s only just the tip of para-linguistics.