Digital Dog Training: Boon or Bust | Cairn Terrier Club of Canada

Have you been bombarded by dramatic on-line advertisements for dog training programs that promise guaranteed results or quick fixes? Did you know that many of these are AI based training programs rather than a real qualified trainer working from safe, reliable and science-based training methodologies to give you accurate training information.

As AI is increasingly taking over the on-line world, it is important to understand the pros and cons of using AI generated information for dog training and to be able to recognize the signs that something may be AI generated.

What formats does AI take in dog training applications?

Smart collars and trackers– designed to monitor movement, location, vital signs, vocalizations etc.
Behaviour analysis apps – these apps use AI supported algorithms to recognize behaviour patterns, document training progress and offer advice
AI supported camera systems to monitor dog movements and alert owners to unusual behaviours.
Digital training – platforms delivering individualized training plans that ideally adjust to progress

While there are both pros and cons associated with the use of smart collars and behaviour analysis apps, the rest of this blog will focus on digital training – the kind of training programs that are increasingly being marketed on social media.

Generic problems with AI

AI is well known to produce incorrect information – so much so that this has been referred to as AI hallucinations. “An AI hallucination occurs when the model produces a response that sounds plausible but is factually incorrect, logically flawed, or completely invented.” AI hallucinations are not a bug, but rather a feature of how large language model AI creates text.
Data bias: AI creates text by scanning data. If the data being scanned is inaccurate, biased, outdated, or slanted, AI will generate text that itself is biased. AI does not reliably evaluate the quality or accuracy of the information it uses to produce its results.

Problems with AI in dog training

Best practices in dog training have evolved over the years as we have developed a deeper understanding of canine cognition and behaviour. Older models of training based on correction and coercion have been shown to be less effective than models based on choice, cooperation and positive reinforcement. Unfortunately, there is still a large body of literature and a significant number of trainers who use, promote, and write about traditional force-based methods – either as a sole method or in combination with positive reinforcement (which produces confusion and conflict in the dog). Unless AI tools are explicitly told to avoid force-based or so called ‘balanced’ methods in their data gathering, AI recommendations will be tainted with what we know to be inappropriate and harmful to dog welfare.
AI based digital training cannot fine tune recommendations based on in-person observations as a qualified trainer does.

What should a dog owner do?

Find a good positive reinforcement-based trainer in your area and work with them. This is especially important if you are having behavioural issues.
When looking for online training materials, seek out known trainers with a strong commitment of positive reinforcement-based training such as
- Donna Hill: https://servicedogtraininginstitute.online/ (more then just SD training)
- Kikkopup: https://www.youtube.com/@kikopup
- George Zak: https://www.youtube.com/zakgeorge21/videos , https://dogtrainingrevolution.podbean.com/
- Susan Garrett: https://www.youtube.com/c/DogsThat/videos
When you are served up ads offering dog training, evaluate carefully to look for signs the content is generated by AI or that training methodology includes techniques based on punishment, coercion or correction.

How to recognize AI generated text

Em dashes: an ‘em dash’ is a dash that is the length of the letter M. To insert an em dash, a human must hold down the shift + option keys while typing a hyphen or select it from the symbol’s menu, therefore most human writers will use quicker and more convenient options such as other punctuation (comma, parenthesis, colon or semicolon) or a hyphen. For some reason, AI loves em dashes. Excessive use oif em dashes is a clue to AI text. For more information on em and en dashes see: https://www.merriam-webster.com/grammar/em-dash-en-dash-how-to-use
Triplet phasing – the human brain is predisposed to remember 3-sets and therefore AI is programmed to describe things in threes. Overuse of triplet phrasing is a common feature of AI writing. g. “Neo doesn’t just load pages, it organizes, summarizes and protects while you browse. It’s like having Chat GPT + Google + Norton security all in one tab.” “no wires, no fuel, no emissions”.
Reversal sentences: Sentences that state what something isn’t (or doesn’t do) then reverse and explains what it is (or does). g. It wasn’t just ‘x’ it was ‘y’. It’s not about xx, it’s about yy. Multiple reversal sentences are a common feature of AI writing.
Highly consistent length of sentence and paragraph. Human writing typically has variation in sentence and paragraph length.
Overuse of dramatic language (content to impress rather than inform)
- attention-grabbing phases or narrative hooks³: Have you ever…., What if I told you….Are you struggling with…. No one saw it coming….
- buzzwords such as game-changing, cutting-edge, revolutionary, innovative, rich tapestry etc.
- capitalization
- emojis
Absence of personal anecdotes or experiences (and personal pronouns)
You can use free AI text detectors to evaluate content, and it will give you a probability of AI generation. https://www.grammarly.com/ai-detectorv

References

Blog